US20120042386A1 - Reputation system for web pages and online entities - Google Patents
Reputation system for web pages and online entities Download PDFInfo
- Publication number
- US20120042386A1 US20120042386A1 US13/196,789 US201113196789A US2012042386A1 US 20120042386 A1 US20120042386 A1 US 20120042386A1 US 201113196789 A US201113196789 A US 201113196789A US 2012042386 A1 US2012042386 A1 US 2012042386A1
- Authority
- US
- United States
- Prior art keywords
- tag
- online entity
- score
- trust
- raters
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Abandoned
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
- G06Q10/107—Computer-aided management of electronic mailing [e-mailing]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Definitions
- Trust is a fundamental quality of individuals interacting in human societies. It is essential for the functioning of many societal institutions, including immigration control, credits and loans, and national security. Thus, a measure of trust would have many applications in these domains and others. Unfortunately, trust has been difficult to measure, and particularly hard to measure in an automated way.
- trust is important for several reasons. First, surveillance is expensive, and thus limiting the number of targets is extremely important. Quantifying the trust associated with each person or entity in a network would be useful to concentrate scarce surveillance resources where they matter most. Second, trust is important for the evaluation of the reliability of tips about suspicious activity. The population at large's eyes and ears would constitute perhaps the best defense against terrorists if there wasn't so many false alarms which currently make following up on tips intractable. The false alarm problem would be much alleviated if a systematic measure of the trust of individual tip providers was available.
- a further problem lies in that we live in an era of information overload.
- Electronic communications constitute a growing fraction of the information that a professional must digest and process daily.
- the boom of low-cost communications is putting tremendous demands on the time of individual professionals.
- Systems to efficiently and automatically prioritize electronic communications are urgently needed.
- the web site repcheck dot corn provides a way to rate the reputation of people over the Internet.
- the system is not particularly user-friendly because it does not employ tags and does not use a tag bar, thus forcing people to go to a third party site to tag and/or read about a person.
- the system does not allow multiple ratings for different tags, confusing high ratings for one reason (e.g. honest) with high ratings for another reason (e.g smart).
- the site only works for people, not pages or sites.
- the web site rapleaf dot corn allows to look people up (by way of their e-mail address) before a purchase or a sale, and rate them afterwards.
- this site does not take into account the rater's reputation.
- the system does not allow multiple ratings for different tags, confusing high ratings for one reason (e.g honest) with high ratings for another reason (e.g. smart).
- people cannot rate and read ratings right from their browser as they are visiting the page.
- only e-mail addresses can be rated. Given that many people do not know the e-mail address of a web site they are considering transacting business with, the usefulness of such method is limited.
- the web site del.icio.us allows users to tag web pages, but these tags are not used as bookmarks to find web pages, nor as descriptors of web pages to be used when visiting the pages, nor do they include any measure of trust or reputation, nor one that takes the tagger's reputation into account.
- the present disclosure addresses the above problems by providing a global measure of trust for each participant in a network as well as a method to calculate it automatically.
- a method for rating online entities, such as online identities is provided, which also takes into account the reputation of the raters.
- a method for rating an online entity in accordance with a tag comprising: arranging the online entity to be rated by a plurality of raters in accordance with the tag; assigning a rater's score to the raters rating the online entity with the tag, to establish reputation of the raters; assigning a tag score to the online entity, the tag score depending on i) ratings provided by the raters to the online entity in accordance with the tag, and ii) the reputation of all raters rating the online entity with the tag.
- a method to establish trust in e-mail communications between senders and receivers comprising: assigning a trust score to an online entity represented by an e-mail address, the trust score depending on: i) amount of communications received by the e-mail address; and ii) a trust indication of all senders of the communications to the e-mail address.
- a method of authenticating an association between a user and a web page comprising: requesting the user to place a code in the web page; searching the code in the web page; and upon finding the code, authenticating the association.
- a method for rating an online entity comprising: arranging the online entity to be rated by a plurality of raters; assigning a rater's score to the raters rating the online entity, to establish reputation of the raters; and assigning a score to the online entity, the score depending on i) ratings provided by the raters to the online entity, and ii) the reputation of all raters rating the online entity, wherein assigning the score to the online entity is established through a raters/entities matrix associated with a graph connecting the raters and the online entity.
- a method of authenticating an association between a user or owner of an e-mail address and a web page comprising: requesting the user or owner of the e-mail address to place a code in the web page; searching the code in the web page; and upon finding the code, authenticating the association.
- a method for rating an online entity comprising: arranging the online entity to be rated by a plurality of raters; assigning a rater's score to the raters rating the online entity, to establish reputation of the raters; and assigning a score to the online entity, the score depending on i) ratings provided by the raters to the online entity, and ii) the reputation of all raters rating the online entity, wherein assigning the score to the online entity is established through a raters/entities matrix associated with a graph connecting the raters and the online entity.
- a toolbar for a web browser comprising tags, description, reviews or ratings assigned to a web page being visited by a user.
- a toolbar for a web browser comprising ads relevant to the web page being visited.
- the trust measure proposed herein is dependent on the network of “endorsements” between members of the network. These endorsements can come in a variety of ways, from explicit declarations of trust between individual members of a networks to implicit measures gleaned from the users' behavior. Different endorsements can signify trust in different contexts.
- the trust measure is global and network-based because the trust assigned to individual components of the network depends on interactions throughout the network. The reason for this is that a node's trust depends not only on who it is trusted by, but also on how trustworthy each of these is. In an aspect of the present disclosure, this gives rise to a recursive formula, whose solution is specified by the present disclosure.
- trust can be used as a proxy for reliability. If a given component X of a system of interdependent parts is relied upon by many other components, and if these other components are reliable, then X must be reliable, too.
- the present disclosure provides a measure for the reliability of components in material systems, such as electronic systems, in addition to those of social networks.
- a tag is a rating language or rating sign (thumbs, stars, cool, trustworthy, scientific, beauty, etc) associated with an online entity.
- An online entity can be a way a physical or commercial person is electronically known through the Internet, e.g. a web page, a web site, an online profile, an e-mail address, a telephone number and so on.
- the online entity can include products (e.g. identified by a bar code, product code, or any other description) or ant tagged or taggable subject.
- FIG. 1 shows a schematic diagram of e-mail communications between senders and receivers.
- FIG. 2 shows a matrix A associated with senders and receivers of the e-mails together with a dominant eigenvector R(A) of matrix A.
- FIG. 3 shows a connectivity graph of communications between senders and receivers.
- FIG. 4 shows a web page rated by raters in accordance with a tag.
- FIG. 5 shows a matrix T associated with a tag.
- the method according to the present disclosure is modeled on reputation in actual human communication networks.
- a human user In order to communicate with trusted members of a network, a human user usually has to first acquire some degree of trust: a bum off the street would find it hard to get a multinational CEO's attention, for example. This system works due to accountability.
- the present disclosure endows individual online identities with a measure of trust. This trust measure is a global function of communications in the network, and does not require a recipient to know a given sender in order to trust communications from said sender. Instead, trust has a transitive property, such that if A trusts B, and B trusts C, then A will trust C to some degree.
- this transitive property does not stop at any particular number of node traversals, but rather follows an infinite recursion.
- this transitive property can be made to decay with each node traversal, making ratings by any node more trusted by a node N the closer they are to that node N, as also later shown in the present disclosure.
- FIG. 1 shows a schematic diagram of e-mail communications between senders and receivers where sender 10 is represented by e-mail address a@b and recipient 20 is represented by e-mail address x@y.
- Each user 10 (whose online identity in this case is an e-mail address a@b) acquires trust points by virtue of receiving correspondence from trusted users S 1 , S 2 , . . . SN.
- a communication 50 is interpreted as a measure of trust awarded from sender S 1 to receiver 10 .
- trust awarded through a communication does not in any way diminish the trust of the sender S 1 , S 2 , . . . SN.
- a sender's trust is divided among all recipients of sender's communications as a monotonous function of the number of communications exchanged. In the simplest such scheme, trust is distributed in proportion to the number of communications exchanged. Trust flows through the system.
- A be a square matrix, each row of which represents a sender (S 1 , S 2 , . . . , SN) and each column of which represents a recipient (R 1 , R 2 , . . . , a@b, . . . ).
- the trust vector sought by the present disclosure is the dominant eigenvector of A, R(A).
- the dominant eigenvector of the matrix is the eigenvector corresponding to the eigenvalue of largest magnitude for that matrix. See, for example, S. Brin, L. Page, “The Anatomy of a Large-Scale Hypertextual Web Search Engine” WWW7/Computer Networks.
- the dominant eigenvector R(A) comprises a plurality of scalar values r(R 1 ), r(R 2 ), r(a@b), . . . , each scalar value indicating the score of a particular recipient (or column of the above described matrix A).
- R(A) comprises a plurality of scalar values r(R 1 ), r(R 2 ), r(a@b), . . . , each scalar value indicating the score of a particular recipient (or column of the above described matrix A).
- the reputation method according to the present disclosure is robust to common manipulations that affect state-of-the-art reputation systems (e.g. EBay's feedback system), such as the creation of numerous identities which endorse each other in an attempt to inflate their trust levels. In particular, until an identity gets endorsements from trusted members of the community, its trust will not increase.
- the system according to the present disclosure allows a gradual increase in the trust conferred on a particular user throughout the user's lifetime, because as the user receives more email from trusted correspondents, his/her trust will increase.
- the rating method can be available to senders S 1 , S 2 , . . . , SN, 20 , for free, allowing widespread adoption of the technology.
- a user's trust rating 30 (see FIG. 1 ) can be transmitted in an encrypted way, preventing tampering or unauthorized reading.
- Recipients 20 (see FIG. 1 ) can choose to buy software to filter out communications that do not reach the desired trust level.
- trust can be computed and transmitted by a third party with access to records of email communications provided by the senders of such communications, to prevent tampering. See related patent application, “Method and System for Authentication of Electronic Communications”, by the same applicant of the present disclosure, already mentioned above, incorporated herein by reference in its entirety. In such embodiments, trust of any node can be ascertained by any client submitting a request to a trust authentication server.
- Recipients can select their desired trust threshold: high thresholds will allow only the most trusted senders' messages in; low thresholds may allow unwanted spam. Trust thresholds may be varied over time, allowing for an increase in thresholds as the system's use becomes widespread. Recipients can also choose to have the trust level of the sender displayed with every incoming communication, as shown in FIG. 1 , allowing rapid prioritization. The person skilled in the art will appreciate that software can be made to automatically prioritize the display of email from senders with higher trust level.
- user 20 's trust of user 10 can also be a function of the distribution of time delays between user 20 's reception of communication from user 10 and user 20 's response, because more important or trusted people may have their emails responded to more quickly (with less lag or delay) than less important or trusted people.
- user 20 's trust of user 10 can also be a function of the distribution of length of messages from user 20 to user 10 as compared to the lengths of messages from user 10 to user 20 , and/or of the ratio computed by dividing the number of messages from user 20 to user 10 over the number of messages from user 10 to user 20 , such that users who get a greater fraction of their outgoing messages responded to are assigned greater trust.
- trust can be calculated in a distributed computation, with a software application in each user's computer calculating that user's trust iteratively, sending only the trust value itself to the outside world.
- the trust of each node N is computed by the information that other nodes convey about the communications that they transmitted to node N. This computation can be performed by each node in the network, or it can be performed by one or more central servers. In this way, no node can attempt to fake its own trust measure.
- trust can be transferred from one identity or email address to another in the case, for example, of a person changing email addresses. See FIG. 1 , where online identity 10 is represented both by e-mail address a@b and by e-mail address c@d.
- a user's trust can be modified by complaints received against the user (e.g. due to spamming) or endorsements. See arrows 60 , 70 shown in FIG. 1 .
- a user's trust can also be modified by the trust of its recipients—in other words, a user pays the price for (or reaps the benefits of) the value of its “endorsements”: recommending untrustworthy users comes at a price, protecting the integrity of the system.
- a user's trust is modified by the derivative of the trust of its recipients following endorsement, rewarding risk-taking “finders” who endorse trustworthy beginners and penalizing recipients who endorse spammers or otherwise untrustworthy users.
- trust conferred on an identity is not a function of the identity alone, but also of a similar function indicating the trust a particular user has of another.
- This embodiment is particularly useful for recommendation systems, since tastes are based on similarities, which are not transitive, and thus an infinitely regressing transitive property is not desired. It may also be desirable in a reputation system, for example where users need to trust individuals who are not globally trustworthy.
- the calculation of trust for a particular user, or recommendation value starts at said user, and value gets diluted with distance from the user, measured in nodes away in the graph of communication connectivity. For example, if user A writes to 10 users with equal frequency, each of those 10 users would have a 10% trust score from A.
- each of those writes to 5 users with equal frequency each of those 5 users would have a 2% trust score from A.
- An example of a graph of communication connectivity is shown in FIG. 3 .
- S 1 has sent 1 communication to R 1 and 2 communications to R 2
- a 1 has sent 1 communication to S 1 and 1 communication to S 2 , and so on.
- trust for each user U is given by a hybrid function combining global trust with a personalized trust function for users close to U, with closeness measured in the sense of distance mentioned above.
- Internet surfers can rate individual websites with arbitrary expressions (e.g. text). For example, a user visiting a web site may rate it with a ‘Thumbs Up’, or with 5 stars, or tag it “cool”, “trustworthy” and “scientific.” Such tagging can be done, for example, via a toolbar for the user's Internet browser.
- FIG. 4 shows a web site 110 rated by raters 130 , 140 , 150 in accordance with a tag 120 .
- raters 170 , 180 , 190 are rated and their reputation or rating 135 , 145 , 155 is taken into account when establishing the rating of the web site 110 .
- each rater 170 , 180 , 190 is associated with a web page 230 , 240 , 250 (also called online profile, web profile, online identity or web identity throughout the present disclosure). This kind of association can also be established by e-mail addresses or profiles uploaded by one or more users with information identifying the user. Therefore, a rater is rated by other raters in the same way a web page not corresponding to a web identity is rated by that rater or the other raters.
- a graph connecting all web pages and raters is established for each tag, similarly to what was already shown in FIG. 3 with reference to the communication embodiment of the present invention.
- a rater/web page ‘tag matrix’ is associated with that graph, as shown in FIG. 5 .
- Rows of the tag matrix denote raters, e.g. 130 , 140 , 150
- columns of the tag matrix denote web pages, e.g. I 1 , I 2 , 110 .
- These web pages sometimes correspond to web pages as such and sometimes correspond to web identities.
- Each cell 300 represents the rating given by rater ‘i’ to web page ‘j’ or rater ‘j’ (i.e. web identity ‘j’).
- This rating can be boolean (tags only, e.g. 1 or 0)—see cells 310 , 320 —or multi-value (e.g. 3 if three stars have been given as a rating)—see cell 330 .
- the reputation for each rated node is given by the dominant eigenvector R(T) of the tag matrix T.
- the dominant eigenvector will comprise a plurality of scalar values r(I 1 ), r(I 2 ), . . . , r( 110 ) . . . , each scalar value indicating the score of a particular page (or column of the above described tag matrix) for that tag.
- the eigenvectors and eigenvalues of the matrix associated with the graph are computed, to establish a metric for each web page/tag combination. For example, the web pages tagged as “cool” by people whose associated webpage/online profile has been rated as “cool” will attain a higher “cool” rating than those web pages which have been tagged “cool” by an equivalent number of people who have lower “cool” ratings. Thus, each page is assigned a score for each tag that it has ever been tagged with.
- a person skilled in the art will know how to calculate a dominant eigenvector, for example through software packages such as Matlab® by Mathworks® or using the iterative algorithm disclosed in the Brin and Page 1998 paper mentioned above. It typically involves computing the dominant eigenvector iteratively by computing the rank of each rated element j as a function of the ranks of the raters who rated j, and then reiterating.
- the amount of rating transferred by a rater can be made proportional to the rater's rank and inversely proportional to the number of ratings given by that rater. Alternatively, it can be made independent of the number of ratings given by the rater.
- the rating of j can be made proportional, in any iteration, to the sum of the amount of rating transferred to it by all raters who rated it.
- a same user can be associated with multiple web pages/rated entities 410 , 420 as shown in FIG. 4 .
- the reputation of each web page 110 , 410 , 420 associated with the user is computed separately, and the reputation for the user is given by the sum of the reputations of each web page associated with him/her.
- the idea behind the sum is that a person's ‘credits and liabilities’ are the sum of his credits and liabilities across all of his ventures and/or personas. Adding one unrated site should not diminish the credibility of Bill Gates.
- the person skilled in the art will understand that other embodiments may be possible, such as the average or indeed any function.
- ratings are given by recommendations provided by members of a social network of others in the network or outside of it.
- ratings are given by who is in a set of top N friends, such as MySpace®.
- ratings are given by who is on a friends list.
- a rating of user B by user A can be given by inclusion of B in any set generated by A explicitly or implicitly through his/her behavior. For example, somebody can be rated as ‘friendly’ just by putting him or her in a list of ‘friends.’
- tags with negative connotations should not weigh more heavily tags from users with that negative tag.
- tags with negative connotation are detected by finding pages which have an overall negative generic rating (more thumbs down than thumbs up), and then weighing the ratings for their tags by using the generic rating of the raters as opposed to the tag-specific ones.
- Generic ratings have a separate kind of matrix. In particular, the generic rating is computed from a matrix with thumbs up/thumbs down ratings associated with no tag.
- tags with negative connotations is to compute how well the tags Y given by a rater with high score for a tag X predicts the score for tag Y computed by the eigenvector of the Y matrix, and use a combination of scores of raters in matrices Y weighted by the predictivities Z (e.g. correlation coefficients).
- the problem with the latter approach is it favors people who score like the masses do, which is not necessarily good.
- a simple transitive case is used, where the tag rank is indicated with TR.
- rating is calculated in view of agreement/disagreement with a tag, e.g. John is a “jolly-good fellow”/John is not a “jolly-good fellow”.
- a universal identifier UI
- identifier ⁇ check mark
- identifier x cross
- users can rate each node using a star scheme or point scale or other rating scheme. Users could give a different rating for each tag.
- the “disagree” ratings influence the tag rank TR.
- TR (node n ,tag t ) sum over all nodes n ′ who have rated n “agree” with tag t of ( TR ( n′,t )/# of nodes n ′ has rated “agree” with tag t ) ⁇ sum over all nodes n ′ who have rated n “disagree” with tag t of ( TR ( n′,t )/# of nodes n ′ has rated “disagree” with tag t ) (Equation 1)
- authoritativeness may be computed by any measure of how well a tag rank for tag t′ predicts accuracy in rating nodes with tag t. For example, the following formula can be used.
- TR (node n ,tag t ) sum over all nodes n ′ who have rated n “agree” with tag t of [sum over all tags t ′ of (predictivity( t′ ⁇ t )* TR ( n′,t ′)/# of nodes n ′ has rated “agree” with tag t /sum over all tags t ′ of predictivity( t′ ⁇ t ))] ⁇ sum over all nodes n ′ who have rated n “disagree” with tag t of [sum over all tags t ′ of (predictivity( t′ ⁇ t )* TR ( n′,t ′)/# of nodes n ′ has rated “disagree” with tag t /sum over all tags t ′ of predictivity( t′ ⁇ t ))] (Equation 2)
- TR for each tag for each node can be displayed as the size of the font of the tag. Negative TRs can cause a tag to be prefixed with the word ‘NOT’.
- tags with transitivity where nodes highly rated for a tag are good judges of others for that tag, e.g. “smart”
- tags with no transitivity e.g. “stupid”
- synonyms see authoritativeness of smart people over intelligence, as explained above
- antonyms and tags with too little data (by pooling in authoritativeness from related tags). It is also robust to manipulation by users voting with the masses, as predictivities are computed for tag pairs, not nodes.
- Some embodiments of the present disclosure can compute certainty of each tag rank, or a confidence interval, or error bar, that expresses how sure one is of the correctness of the value. As the amount of data used to compute a tag rank increases, the certainty increases.
- TR certainty TRC
- the following formula can be used:
- TRC ( TR ( n,t )) square root of the sum over all nodes n ′ who have rated node n (agree or disagree) with tag t and over all tags t ′ of (predictivity( t′ ⁇ t )* TR ( n′,t ′)/# of nodes n ′ has rated “agree” with tag t ) (Equation 3)
- TRCs can be converted into the percentile of TRC (n,t) with respect to all TRCs for tag t, i.e. the percentage of nodes with a higher or lower tag rank certainty for that tag. This percentile could also be computed with respect to all TRCs for node n if one is to express how common different tags are for a node without normalizing for overall tag frequency. TRCs for each tag for each node can be displayed as the contrast of the font of the tag. Tag ranks can also be expressed as percentiles.
- a problem may arise with the use of many tags, which is that reputation information might lie distributed across the tag matrices corresponding to multiple tags with related meanings. For example, some people may label a node as funny while others label it as humorous.
- this problem is addressed in the present disclosure by computing the probability that one tag predicts another, or correlation between tags, and computing a modified score for tag X that is the weighted average of the score for each Tag Y multiplied by the correlation between Y and X, or p(X
- a correlation between tags is defined as follows: given two vectors containing the probability of each rated node getting tagged with given tags X and Y, respectively, one embodiment of the correlation between these two tags is given by the correlation coefficient between the two dominant eigenvectors of X and Y. If the tags are related, the sites will obtain similar scores for both tags.
- a further problem may arise when not enough data is available to construct a reliable tag matrix for a given tag, even after employing the above discussed correlation.
- One way in which the present disclosure addresses this problem is by using calculation of a score through the dominant eigenvector for a generic tagless rating matrix (employing a generic thumbs up/thumbs down or ratings system for each node) for node/tag pairs if they do not have enough data to compute a score for each tag, in combination with the actual tags applied to the node. Therefore, the rating of a generally highly rated rater who rates a node with a new tag would get weighted highly until enough data exists for the graph of that tag to evaluate tag-specific ratings for each rater.
- the same equations used above can be used in this instance, allowing nodes ranked with one tag to convey their influence to nodes ranked with another related tag.
- the method in accordance with the present disclosure gives higher values for older nodes, because a new node cannot be rated by an older one that is no longer in use. For example, while Bill Clinton can rate Winston Churchill as a great leader, Churchill cannot return the favor.
- This inconsistency can be solved by keeping the dates of each rating and dividing the raw score of node X, obtained through the method of the present disclosure, by the fraction of all ratings/hyperlinks that are newer than the age of node X.
- Another embodiment divides by the number of nodes that are newer than node X, but the former is preferred.
- All or the top tags for each web page can then be displayed for each web page and/or online identity.
- they are displayed in a toolbar in the Internet browser, see toolbar 200 of FIG. 4 .
- they are displayed in a web page 450 associated with each web page 110 .
- they are displayed in a display superimposed on the web page. In this case, the display can be triggered by a browser plugin, for example.
- FIG. 4 shows an example of a toolbar 200 on a web browser.
- a toolbar is a frame or a special area of a web browser than contains icons allowing commands to be executed (tools, menu items, actions) and/or information.
- Toolbar 200 comprises, for example, a field 120 for a tag name to be assigned by the user, a symbol 510 (which can either be positive or negative or neither, such as a thumbs up or a thumbs down symbol), a score 160 , and other additional fields 520 , 530 such as a ‘more info’ field and a ‘my stats’ field.
- the toolbar 200 can also display ads and/or unpaid links to web pages, news and/or documents related to the web page 110 being browsed or the user's activity with any document in their computer to be shown, e.g., in fields 520 , 530 . Similarity between the web page 110 being browsed and the web pages to which ads or links lead to can be assayed by several methods familiar to one skilled in the art, including link analysis, semantic analysis, user behavior etc. Alternatively, a link can be offered to coupons for the web site/profile in question on the toolbar when viewing the corresponding web page.
- the present disclosure also includes a toolbar for an Internet browser, similar in aspect to toolbar 200 discussed above, that provides information on the URL being visited and/or a link to such information.
- information could contain, for example, ads (i.e. advertisings) relevant to the web page being visited.
- ads i.e. advertisings
- Relevance of ads to a web page can be computed by any measure of semantic similarity between representations of the ad or the web page that clicking on the ad leads to, and of the web page being visited, for example, by latent semantic analysis (LSA) or by the dot product between vectors of word occurrences.
- LSA latent semantic analysis
- Relevance can also be personalized to take into account the probability, based on a user's past behavior (including web pages visited, for example), that the user would be interested in an ad.
- a possible field of use of the methods in accordance with the present disclosure is web search. If it is known what pages rank highly for a tag, this can be used to return results during the course of a tag search or a related query. In other words, users can search ‘tagged’ pages on the Web using keywords and filter web pages to be shown only those with a rating above a threshold for a particular tag (e.g. “cool” or “trustworthy”).
- a generic “tagless” rating can be used for generic rankings of web pages and/or people, and used when raters emit a rating (e.g. thumbs up/thumbs down without typing in a tag or description.
- a rating e.g. thumbs up/thumbs down without typing in a tag or description.
- the raters' tagless or generic rating or reputation is used to weigh their ratings of other people and/or web pages regardless of the tag they employ.
- a web page such as an online profile of a person
- One instance of the problem reduces to that of associating an email address with a web page.
- this is accomplished by providing the person who wishes to verify that they are associated with a given web page with a code (for example a randomly generated string of text and/or numbers and/or other symbols) that the person must insert in the (code for the) corresponding web page.
- a code for example a randomly generated string of text and/or numbers and/or other symbols
- the person notifies a service (e.g. clicks on a link) and a server or client verifies whether the code has indeed been inserted in the corresponding web page.
- a service e.g. clicks on a link
- a server or client verifies whether the code has indeed been inserted in the corresponding web page.
- the association between person and web page gets established. After such event, the person will be free to remove the code from the web page, if it is so desired.
- the tags and reputation assigned to the web page can subsequently be transferred to the person and affect his/her ratings of other web pages or profiles.
- ownership or control of the web page is verified by forcing the owner of the site to place the code in a non-comment section of the web page that is not open for all to modify before the association between web page and user is established.
- the user can be the owner of an email address. Further, the user can be identified by a user identifier, such as a username/password combination or an email address.
- the present disclosure provides a method by which it is possible to automatically detect which tags share a ‘rater to rating transitivity’ (such as the ‘trustworthiness’ rating) and which do not (such as the ‘beauty’ rating).
- One way of doing that is to measure the correlation between raters' ratings for tag T and the correlation between a rater's rating of web pages or profiles X and the computed score for the corresponding tag ranks of X in a first iteration of the tag rank computation, before the tag rank of a rater influences their influence.
- raters highly rated for that tag are more predictive of the score than raters lowly rated for that tag and should thus be counted more heavily, so the standard tag rank algorithm can be applied.
- Such tags are said to exhibit rater to rating transitivity (RTRT) and for them, the score computation is performed as described above.
- users of a recommendation network can be allowed to find out how reputable each member in the network is without needing to know how trustworthy each of the recommenders is. This is done by computing, for example, the principal Eigenvector of a recommendation network made up of recommendations or endorsements between individuals, to compute the reputation of each individual.
- the ratings can also come in the form of communications. For example, if member A sends an e-mail to member B, that could be counted as a vote from A to B. Likewise, if A replies fast to an e-mail from B, that can count as a vote from A to B inversely proportional to (or monotonically decreasing with) the time delay it took A to respond since the e-mail from B was received. Moreover, the use of reputation measured as the eigenvectors in a communication matrix of such communication can be used to combat spam.
- the reputation of any node N in a network or graph with edges representing positive and negative ‘endorsements’, ratings or reviews can be compared as a function of the corresponding scalar from the dominant eigenvector of the matrix of positive endorsements, the number of negative endorsements received by node N, the number of negative endorsements given by each of the nodes that gave N a negative endorsement, and the reputation of the nodes that gave N a negative endorsement.
- the reputation of node N can be given by Equation 1 above.
- reputation of a user can be visible when the user leaves a comment on a web page of a web site (e.g. a blog).
- the comment can include computer program code that checks a server for reputation scores and/or tags when a web browser fetches the web page, so that the reputation is constantly updated. Reviews or comments left by the user will be more useful to other users if they know whether that user is trustworthy or not.
- the tag scores can be personalized by counting ratings by people with tastes similar to each user more than others.
- similarity is a measure of the similarity in ratings between user i and user j.
- similarity is the cosine of the angle between, or the dot product of, the vectors representing users i and j in a matrix which is composed of the N principal components of a matrix M whose rows represent users and whose columns represent nodes (taggable objects) and where each cell contains a user's rating of a taggable object.
- a singular value decomposition is performed to compute the N principal components of M using principal components analysis (PCA), project each user's row vector into the reduced dimensionality space, and measure their similarity between the projections of two user-vectors as their dot product or the cosine of the angle between them.
- PCA principal components analysis
- the cosine is usually preferable as it can be negative and can thus express users with opposite tastes, and it is normalized to the interval between ⁇ 1 and 1.
- Equation 6 also includes a normalization, i.e. division by the sum over all ratings by j with tag t.
- the sum of Equation 2 includes positive terms in its first portion and negative terms in its second portion.
- the trust that user A should have in user B is given also by how much “dilution” has happened between A and B, where dilution is such that each node endorsed by A has 1/n of the trust of A, with n the number of nodes endorsed by A, and each node endorsed by a node X endorsed by A has 1/m of the trust that A has in X, and so on.
- the non-global trust can apply to more than just email networks.
Abstract
A method for providing a measure of trust for each participant in a network is disclosed, together with a method to calculate it automatically. In particular, a method for rating online entities, such as online identities is provided, which also takes into account the reputation of the raters.
Description
- The present application is a continuation application of U.S. patent application Ser. No. 11/704,730 filed on Feb. 9, 2007 which claims the benefit of U.S. Provisional Application No. 60/772,063 filed on Feb. 9, 2006 and U.S. Provisional Application No. 60/772,301 filed on Feb. 9, 2006, the contents of both of which are incorporated by reference herein. The present application may be related to U.S. patent application Ser. No. 11/704,729 filed on Feb. 9, 2007 entitled “Method and System for Authentication of Electronic Communications”, also incorporated by reference in the present application.
- This invention was made in part with government support under a MICS Lab CFD, Grid and Other AMD grant awarded by the Office of Science of the DOE. The U.S. Government may have certain rights in the invention.
- The present disclosure relates to the Internet and networks in general. In particular, it relates to a reputation system for web pages and online entities.
- Currently, Internet surfers or, more generally, participants in a network, have little way to know the reputation of a given web page, web site or person represented by an online entity (e.g. in a social network, to distinguish trusted people from would-be predators).
- Trust is a fundamental quality of individuals interacting in human societies. It is essential for the functioning of many societal institutions, including immigration control, credits and loans, and national security. Thus, a measure of trust would have many applications in these domains and others. Unfortunately, trust has been difficult to measure, and particularly hard to measure in an automated way.
- In a security context, trust is important for several reasons. First, surveillance is expensive, and thus limiting the number of targets is extremely important. Quantifying the trust associated with each person or entity in a network would be useful to concentrate scarce surveillance resources where they matter most. Second, trust is important for the evaluation of the reliability of tips about suspicious activity. The population at large's eyes and ears would constitute perhaps the best defense against terrorists if there weren't so many false alarms which currently make following up on tips intractable. The false alarm problem would be much alleviated if a systematic measure of the trust of individual tip providers was available.
- Spam, or electronic communications unwanted by recipients, is a growing problem, to the point that many users receive more spam than desirable email. The problem is even driving pushes for new legislation. Because email travels along global communication networks, though, solutions at the national level are unlikely to succeed. Although anti-spam software exists, spammers are continually modifying the text of their messages, in an evolutionary race to beat anti-spam software. Inevitably, anti-spam software that relies on text filtering is one step behind spam.
- There are two main reasons why it is so hard to enforce spam control. One is that senders regularly fake the sender's email address. This problem is addressed by related patent application. “Method and System for Authentication of Electronic Communications”, by the same applicant of the present disclosure, U.S. patent application Ser. No. 11/704,729 filed on Feb. 9, 2007, which is incorporated herein by reference in its entirety. The second reason is that spammers regularly obtain valid email addresses of their own, which they would be able to authenticate as their legal owners, but which exist only for brief periods of time, for the sole purpose of sending one spam campaign.
- A further problem lies in that we live in an era of information overload. Electronic communications constitute a growing fraction of the information that a professional must digest and process daily. The boom of low-cost communications is putting tremendous demands on the time of individual professionals. Systems to efficiently and automatically prioritize electronic communications are urgently needed.
- The web site repcheck dot corn provides a way to rate the reputation of people over the Internet. However, the system is not particularly user-friendly because it does not employ tags and does not use a tag bar, thus forcing people to go to a third party site to tag and/or read about a person. Moreover, the system does not allow multiple ratings for different tags, confusing high ratings for one reason (e.g. honest) with high ratings for another reason (e.g smart). Further, the site only works for people, not pages or sites.
- The web site rapleaf dot corn allows to look people up (by way of their e-mail address) before a purchase or a sale, and rate them afterwards. However, this site does not take into account the rater's reputation. Moreover, the system does not allow multiple ratings for different tags, confusing high ratings for one reason (e.g honest) with high ratings for another reason (e.g. smart). In addition, people cannot rate and read ratings right from their browser as they are visiting the page. Further, only e-mail addresses can be rated. Given that many people do not know the e-mail address of a web site they are considering transacting business with, the usefulness of such method is limited.
- The web site del.icio.us allows users to tag web pages, but these tags are not used as bookmarks to find web pages, nor as descriptors of web pages to be used when visiting the pages, nor do they include any measure of trust or reputation, nor one that takes the tagger's reputation into account.
- The present disclosure addresses the above problems by providing a global measure of trust for each participant in a network as well as a method to calculate it automatically. In particular, a method for rating online entities, such as online identities, is provided, which also takes into account the reputation of the raters.
- According to a first embodiment, a method for rating an online entity in accordance with a tag is provided, the method comprising: arranging the online entity to be rated by a plurality of raters in accordance with the tag; assigning a rater's score to the raters rating the online entity with the tag, to establish reputation of the raters; assigning a tag score to the online entity, the tag score depending on i) ratings provided by the raters to the online entity in accordance with the tag, and ii) the reputation of all raters rating the online entity with the tag.
- According to a second embodiment, a method to establish trust in e-mail communications between senders and receivers is provided, comprising: assigning a trust score to an online entity represented by an e-mail address, the trust score depending on: i) amount of communications received by the e-mail address; and ii) a trust indication of all senders of the communications to the e-mail address.
- According to a third embodiment, a method of authenticating an association between a user and a web page is provided, comprising: requesting the user to place a code in the web page; searching the code in the web page; and upon finding the code, authenticating the association.
- According to a fourth embodiment, a method for rating an online entity is provided, the method comprising: arranging the online entity to be rated by a plurality of raters; assigning a rater's score to the raters rating the online entity, to establish reputation of the raters; and assigning a score to the online entity, the score depending on i) ratings provided by the raters to the online entity, and ii) the reputation of all raters rating the online entity, wherein assigning the score to the online entity is established through a raters/entities matrix associated with a graph connecting the raters and the online entity.
- According to a fifth embodiment, a method of authenticating an association between a user or owner of an e-mail address and a web page is provided, comprising: requesting the user or owner of the e-mail address to place a code in the web page; searching the code in the web page; and upon finding the code, authenticating the association.
- According to a sixth embodiment, a method for rating an online entity is disclosed, the method comprising: arranging the online entity to be rated by a plurality of raters; assigning a rater's score to the raters rating the online entity, to establish reputation of the raters; and assigning a score to the online entity, the score depending on i) ratings provided by the raters to the online entity, and ii) the reputation of all raters rating the online entity, wherein assigning the score to the online entity is established through a raters/entities matrix associated with a graph connecting the raters and the online entity.
- According to a seventh embodiment, a toolbar for a web browser is disclosed, the toolbar comprising tags, description, reviews or ratings assigned to a web page being visited by a user.
- According to an eighth embodiment, a toolbar for a web browser is provided, the toolbar comprising ads relevant to the web page being visited.
- Further embodiments are present throughout the specification and in the claims of the present application.
- The trust measure proposed herein is dependent on the network of “endorsements” between members of the network. These endorsements can come in a variety of ways, from explicit declarations of trust between individual members of a networks to implicit measures gleaned from the users' behavior. Different endorsements can signify trust in different contexts.
- The trust measure is global and network-based because the trust assigned to individual components of the network depends on interactions throughout the network. The reason for this is that a node's trust depends not only on who it is trusted by, but also on how trustworthy each of these is. In an aspect of the present disclosure, this gives rise to a recursive formula, whose solution is specified by the present disclosure.
- In accordance with the present disclosure, trust can be used as a proxy for reliability. If a given component X of a system of interdependent parts is relied upon by many other components, and if these other components are reliable, then X must be reliable, too. Thus, the present disclosure provides a measure for the reliability of components in material systems, such as electronic systems, in addition to those of social networks.
- As used in the present disclosure, a tag is a rating language or rating sign (thumbs, stars, cool, trustworthy, scientific, beauty, etc) associated with an online entity. An online entity can be a way a physical or commercial person is electronically known through the Internet, e.g. a web page, a web site, an online profile, an e-mail address, a telephone number and so on. The online entity can include products (e.g. identified by a bar code, product code, or any other description) or ant tagged or taggable subject.
-
FIG. 1 shows a schematic diagram of e-mail communications between senders and receivers. -
FIG. 2 shows a matrix A associated with senders and receivers of the e-mails together with a dominant eigenvector R(A) of matrix A. -
FIG. 3 shows a connectivity graph of communications between senders and receivers. -
FIG. 4 shows a web page rated by raters in accordance with a tag. -
FIG. 5 shows a matrix T associated with a tag. - The method according to the present disclosure is modeled on reputation in actual human communication networks. In order to communicate with trusted members of a network, a human user usually has to first acquire some degree of trust: a bum off the street would find it hard to get a multinational CEO's attention, for example. This system works due to accountability. The present disclosure endows individual online identities with a measure of trust. This trust measure is a global function of communications in the network, and does not require a recipient to know a given sender in order to trust communications from said sender. Instead, trust has a transitive property, such that if A trusts B, and B trusts C, then A will trust C to some degree. For global reputation of scoring, this transitive property does not stop at any particular number of node traversals, but rather follows an infinite recursion. For local reputation scoring, this transitive property can be made to decay with each node traversal, making ratings by any node more trusted by a node N the closer they are to that node N, as also later shown in the present disclosure.
-
FIG. 1 shows a schematic diagram of e-mail communications between senders and receivers wheresender 10 is represented by e-mail address a@b andrecipient 20 is represented by e-mail address x@y. Each user 10 (whose online identity in this case is an e-mail address a@b) acquires trust points by virtue of receiving correspondence from trusted users S1, S2, . . . SN. In other words, a communication 50 is interpreted as a measure of trust awarded from sender S1 toreceiver 10. Note that trust awarded through a communication does not in any way diminish the trust of the sender S1, S2, . . . SN. A sender's trust is divided among all recipients of sender's communications as a monotonous function of the number of communications exchanged. In the simplest such scheme, trust is distributed in proportion to the number of communications exchanged. Trust flows through the system. - With reference to
FIG. 2 , let A be a square matrix, each row of which represents a sender (S1, S2, . . . , SN) and each column of which represents a recipient (R1, R2, . . . , a@b, . . . ). The trust vector sought by the present disclosure is the dominant eigenvector of A, R(A). The dominant eigenvector of the matrix is the eigenvector corresponding to the eigenvalue of largest magnitude for that matrix. See, for example, S. Brin, L. Page, “The Anatomy of a Large-Scale Hypertextual Web Search Engine” WWW7/Computer Networks. The dominant eigenvector R(A) comprises a plurality of scalar values r(R1), r(R2), r(a@b), . . . , each scalar value indicating the score of a particular recipient (or column of the above described matrix A). Of course, a similar result can be obtained by exchanging rows with columns. - The reputation method according to the present disclosure is robust to common manipulations that affect state-of-the-art reputation systems (e.g. EBay's feedback system), such as the creation of numerous identities which endorse each other in an attempt to inflate their trust levels. In particular, until an identity gets endorsements from trusted members of the community, its trust will not increase. The system according to the present disclosure allows a gradual increase in the trust conferred on a particular user throughout the user's lifetime, because as the user receives more email from trusted correspondents, his/her trust will increase.
- The rating method can be available to senders S1, S2, . . . , SN, 20, for free, allowing widespread adoption of the technology. A user's trust rating 30 (see
FIG. 1 ) can be transmitted in an encrypted way, preventing tampering or unauthorized reading. Recipients 20 (seeFIG. 1 ) can choose to buy software to filter out communications that do not reach the desired trust level. In some embodiments, trust can be computed and transmitted by a third party with access to records of email communications provided by the senders of such communications, to prevent tampering. See related patent application, “Method and System for Authentication of Electronic Communications”, by the same applicant of the present disclosure, already mentioned above, incorporated herein by reference in its entirety. In such embodiments, trust of any node can be ascertained by any client submitting a request to a trust authentication server. - Recipients can select their desired trust threshold: high thresholds will allow only the most trusted senders' messages in; low thresholds may allow unwanted spam. Trust thresholds may be varied over time, allowing for an increase in thresholds as the system's use becomes widespread. Recipients can also choose to have the trust level of the sender displayed with every incoming communication, as shown in
FIG. 1 , allowing rapid prioritization. The person skilled in the art will appreciate that software can be made to automatically prioritize the display of email from senders with higher trust level. - In one embodiment of the present disclosure,
user 20's trust of user 10 (seeFIG. 1 ) can also be a function of the distribution of time delays betweenuser 20's reception of communication fromuser 10 anduser 20's response, because more important or trusted people may have their emails responded to more quickly (with less lag or delay) than less important or trusted people. - In a further embodiment,
user 20's trust ofuser 10 can also be a function of the distribution of length of messages fromuser 20 touser 10 as compared to the lengths of messages fromuser 10 touser 20, and/or of the ratio computed by dividing the number of messages fromuser 20 touser 10 over the number of messages fromuser 10 touser 20, such that users who get a greater fraction of their outgoing messages responded to are assigned greater trust. - In still a further embodiment, trust can be calculated in a distributed computation, with a software application in each user's computer calculating that user's trust iteratively, sending only the trust value itself to the outside world. In some embodiments, the trust of each node N is computed by the information that other nodes convey about the communications that they transmitted to node N. This computation can be performed by each node in the network, or it can be performed by one or more central servers. In this way, no node can attempt to fake its own trust measure.
- In another embodiment, trust can be transferred from one identity or email address to another in the case, for example, of a person changing email addresses. See
FIG. 1 , whereonline identity 10 is represented both by e-mail address a@b and by e-mail address c@d. - In still another embodiment of the invention, a user's trust can be modified by complaints received against the user (e.g. due to spamming) or endorsements. See
arrows FIG. 1 . In another embodiment, a user's trust can also be modified by the trust of its recipients—in other words, a user pays the price for (or reaps the benefits of) the value of its “endorsements”: recommending untrustworthy users comes at a price, protecting the integrity of the system. In another embodiment, a user's trust is modified by the derivative of the trust of its recipients following endorsement, rewarding risk-taking “finders” who endorse trustworthy beginners and penalizing recipients who endorse spammers or otherwise untrustworthy users. - In an alternative embodiment of the present disclosure, trust conferred on an identity is not a function of the identity alone, but also of a similar function indicating the trust a particular user has of another. This embodiment is particularly useful for recommendation systems, since tastes are based on similarities, which are not transitive, and thus an infinitely regressing transitive property is not desired. It may also be desirable in a reputation system, for example where users need to trust individuals who are not globally trustworthy. In this embodiment, the calculation of trust for a particular user, or recommendation value, starts at said user, and value gets diluted with distance from the user, measured in nodes away in the graph of communication connectivity. For example, if user A writes to 10 users with equal frequency, each of those 10 users would have a 10% trust score from A. If each of those writes to 5 users with equal frequency, each of those 5 users would have a 2% trust score from A. An example of a graph of communication connectivity is shown in
FIG. 3 . S1 has sent 1 communication to R1 and 2 communications to R2, A1 has sent 1 communication to S1 and 1 communication to S2, and so on. - In another embodiment, trust for each user U is given by a hybrid function combining global trust with a personalized trust function for users close to U, with closeness measured in the sense of distance mentioned above.
- Rating of Online Identities in Accordance with a Tag
- In accordance with an additional embodiment of the present disclosure, Internet surfers can rate individual websites with arbitrary expressions (e.g. text). For example, a user visiting a web site may rate it with a ‘Thumbs Up’, or with 5 stars, or tag it “cool”, “trustworthy” and “scientific.” Such tagging can be done, for example, via a toolbar for the user's Internet browser.
FIG. 4 shows aweb site 110 rated byraters tag 120. - Also raters 170, 180, 190 are rated and their reputation or
rating web site 110. In order to build a rater's reputation, each rater 170, 180, 190 is associated with aweb page - A graph connecting all web pages and raters is established for each tag, similarly to what was already shown in
FIG. 3 with reference to the communication embodiment of the present invention. A rater/web page ‘tag matrix’ is associated with that graph, as shown inFIG. 5 . Rows of the tag matrix denote raters, e.g. 130, 140, 150, and columns of the tag matrix denote web pages, e.g. I1, I2, 110. These web pages sometimes correspond to web pages as such and sometimes correspond to web identities. Eachcell 300 represents the rating given by rater ‘i’ to web page ‘j’ or rater ‘j’ (i.e. web identity ‘j’). The person skilled in the art will understand that a similar construction can be obtained by exchanging rows with columns. - This rating can be boolean (tags only, e.g. 1 or 0)—see
cells 310, 320—or multi-value (e.g. 3 if three stars have been given as a rating)—seecell 330. Similarly to what discussed in the e-mail embodiment ofFIGS. 1-3 , the reputation for each rated node is given by the dominant eigenvector R(T) of the tag matrix T. The dominant eigenvector will comprise a plurality of scalar values r(I1), r(I2), . . . , r(110) . . . , each scalar value indicating the score of a particular page (or column of the above described tag matrix) for that tag. - The eigenvectors and eigenvalues of the matrix associated with the graph are computed, to establish a metric for each web page/tag combination. For example, the web pages tagged as “cool” by people whose associated webpage/online profile has been rated as “cool” will attain a higher “cool” rating than those web pages which have been tagged “cool” by an equivalent number of people who have lower “cool” ratings. Thus, each page is assigned a score for each tag that it has ever been tagged with.
- A person skilled in the art will know how to calculate a dominant eigenvector, for example through software packages such as Matlab® by Mathworks® or using the iterative algorithm disclosed in the Brin and Page 1998 paper mentioned above. It typically involves computing the dominant eigenvector iteratively by computing the rank of each rated element j as a function of the ranks of the raters who rated j, and then reiterating. The amount of rating transferred by a rater can be made proportional to the rater's rank and inversely proportional to the number of ratings given by that rater. Alternatively, it can be made independent of the number of ratings given by the rater. The rating of j can be made proportional, in any iteration, to the sum of the amount of rating transferred to it by all raters who rated it.
- In a further embodiment of the present disclosure, a same user can be associated with multiple web pages/rated
entities FIG. 4 . Should this be the case, the reputation of eachweb page - Also, in some embodiments, ratings are given by recommendations provided by members of a social network of others in the network or outside of it. In some embodiments, ratings are given by who is in a set of top N friends, such as MySpace®. In other embodiments, ratings are given by who is on a friends list. In general, a rating of user B by user A can be given by inclusion of B in any set generated by A explicitly or implicitly through his/her behavior. For example, somebody can be rated as ‘friendly’ just by putting him or her in a list of ‘friends.’
- The reputation for tags with negative connotations (e.g. untrustworthy) should not weigh more heavily tags from users with that negative tag. According to an embodiment of the present disclosure, tags with negative connotation are detected by finding pages which have an overall negative generic rating (more thumbs down than thumbs up), and then weighing the ratings for their tags by using the generic rating of the raters as opposed to the tag-specific ones. Generic ratings have a separate kind of matrix. In particular, the generic rating is computed from a matrix with thumbs up/thumbs down ratings associated with no tag.
- An alternative embodiment for tags with negative connotations is to compute how well the tags Y given by a rater with high score for a tag X predicts the score for tag Y computed by the eigenvector of the Y matrix, and use a combination of scores of raters in matrices Y weighted by the predictivities Z (e.g. correlation coefficients). However, the problem with the latter approach is it favors people who score like the masses do, which is not necessarily good.
- In the following paragraphs some mathematical formulations will be provided. As a starting point, a simple transitive case is used, where the tag rank is indicated with TR. In this example, rating is calculated in view of agreement/disagreement with a tag, e.g. John is a “jolly-good fellow”/John is not a “jolly-good fellow”. By way of example, a universal identifier (UI) could be provided to allow raters to easily agree/disagree with common tags previously given to the entity being viewed. For example, identifier √ (check mark) could be provided for agreement and identifier x (cross) for disagreement. Alternatively, an identifier commonly used by a community or by a particular rater could be provided. Alternatively, users can rate each node using a star scheme or point scale or other rating scheme. Users could give a different rating for each tag.
- In the following example, the “disagree” ratings influence the tag rank TR.
-
TR(node n,tag t)=sum over all nodes n′ who have rated n “agree” with tag t of (TR(n′,t)/# of nodes n′ has rated “agree” with tag t)−sum over all nodes n′ who have rated n “disagree” with tag t of (TR(n′,t)/# of nodes n′ has rated “disagree” with tag t) (Equation 1) - In the general case, one may compute the tag rank of a node N for tag t taking into account not only the tag rank of the raters of N for tag t, but also their tag rank for other tags for which a tag rank denotes authoritativeness on tag t. For example, someone with a high tag rank for the tag “smart” will have a high influence for the purpose of tagging “intelligent”, not just “smart.” In particular, authoritativeness may be computed by any measure of how well a tag rank for tag t′ predicts accuracy in rating nodes with tag t. For example, the following formula can be used.
-
TR(node n,tag t)=sum over all nodes n′ who have rated n “agree” with tag t of [sum over all tags t′ of (predictivity(t′→t)*TR(n′,t′)/# of nodes n′ has rated “agree” with tag t/sum over all tags t′ of predictivity(t′→t))]−sum over all nodes n′ who have rated n “disagree” with tag t of [sum over all tags t′ of (predictivity(t′→t)*TR(n′,t′)/# of nodes n′ has rated “disagree” with tag t/sum over all tags t′ of predictivity(t′→t))] (Equation 2) - Where
-
- predictivity(t′→*t)=coefficient of correlation between TR(t′) and prediction accuracy, computed over all nodes (predictivity goes between −1 and 1)
- prediction accuracy=1-prediction error
- prediction error (t′,t,node n)=abs(rating(t,n′,n)−TR(t,n′)) (this assumes that TR's go between 0 and 1, −1 and 1 after the subtraction of the disagrees)
- rating(t,n′,n) is the rating for tag t given by n to n′
- a rating is 1 if agree and −1 if disagree
- TR for each tag for each node can be displayed as the size of the font of the tag. Negative TRs can cause a tag to be prefixed with the word ‘NOT’.
- This definition takes care of tags with transitivity (where nodes highly rated for a tag are good judges of others for that tag, e.g. “smart”), tags with no transitivity (e.g. “stupid”), synonyms (see authoritativeness of smart people over intelligence, as explained above), antonyms—and tags with too little data (by pooling in authoritativeness from related tags). It is also robust to manipulation by users voting with the masses, as predictivities are computed for tag pairs, not nodes.
- Some embodiments of the present disclosure can compute certainty of each tag rank, or a confidence interval, or error bar, that expresses how sure one is of the correctness of the value. As the amount of data used to compute a tag rank increases, the certainty increases. As for TR certainty (TRC), the following formula can be used:
-
TRC(TR(n,t))=square root of the sum over all nodes n′ who have rated node n (agree or disagree) with tag t and over all tags t′ of (predictivity(t′→t)*TR(n′,t′)/# of nodes n′ has rated “agree” with tag t) (Equation 3) - The presence of the square root is due to the fact that the error goes up with square root of the number of samples. The rest of the expression is a generalization of the number of samples, where each rating sample is weighed by its author's TR for tag t′ and by the predictivity(t′→t), but inversely by how promiscuous or generous the rater is with tag t. Any measure that increases with the amount of data that went into the computation of a tag rank or decreases with the expected sample error will do, though.
- For the purpose of display, TRCs can be converted into the percentile of TRC (n,t) with respect to all TRCs for tag t, i.e. the percentage of nodes with a higher or lower tag rank certainty for that tag. This percentile could also be computed with respect to all TRCs for node n if one is to express how common different tags are for a node without normalizing for overall tag frequency. TRCs for each tag for each node can be displayed as the contrast of the font of the tag. Tag ranks can also be expressed as percentiles.
- A problem may arise with the use of many tags, which is that reputation information might lie distributed across the tag matrices corresponding to multiple tags with related meanings. For example, some people may label a node as funny while others label it as humorous. In some embodiments, this problem is addressed in the present disclosure by computing the probability that one tag predicts another, or correlation between tags, and computing a modified score for tag X that is the weighted average of the score for each Tag Y multiplied by the correlation between Y and X, or p(X|Y). A correlation between tags is defined as follows: given two vectors containing the probability of each rated node getting tagged with given tags X and Y, respectively, one embodiment of the correlation between these two tags is given by the correlation coefficient between the two dominant eigenvectors of X and Y. If the tags are related, the sites will obtain similar scores for both tags.
- A further problem may arise when not enough data is available to construct a reliable tag matrix for a given tag, even after employing the above discussed correlation. One way in which the present disclosure addresses this problem is by using calculation of a score through the dominant eigenvector for a generic tagless rating matrix (employing a generic thumbs up/thumbs down or ratings system for each node) for node/tag pairs if they do not have enough data to compute a score for each tag, in combination with the actual tags applied to the node. Therefore, the rating of a generally highly rated rater who rates a node with a new tag would get weighted highly until enough data exists for the graph of that tag to evaluate tag-specific ratings for each rater. Alternatively, the same equations used above can be used in this instance, allowing nodes ranked with one tag to convey their influence to nodes ranked with another related tag.
- The method in accordance with the present disclosure gives higher values for older nodes, because a new node cannot be rated by an older one that is no longer in use. For example, while Bill Clinton can rate Winston Churchill as a great leader, Churchill cannot return the favor. This inconsistency can be solved by keeping the dates of each rating and dividing the raw score of node X, obtained through the method of the present disclosure, by the fraction of all ratings/hyperlinks that are newer than the age of node X. Another embodiment divides by the number of nodes that are newer than node X, but the former is preferred.
- All or the top tags for each web page, as ranked by tag rank, tag rank certainty or some combination of the two, can then be displayed for each web page and/or online identity. In one embodiment, they are displayed in a toolbar in the Internet browser, see
toolbar 200 ofFIG. 4 . In another embodiment, they are displayed in aweb page 450 associated with eachweb page 110. In a further embodiment, they are displayed in a display superimposed on the web page. In this case, the display can be triggered by a browser plugin, for example. -
FIG. 4 shows an example of atoolbar 200 on a web browser. A toolbar is a frame or a special area of a web browser than contains icons allowing commands to be executed (tools, menu items, actions) and/or information.Toolbar 200 comprises, for example, afield 120 for a tag name to be assigned by the user, a symbol 510 (which can either be positive or negative or neither, such as a thumbs up or a thumbs down symbol), ascore 160, and otheradditional fields toolbar 200 can also display ads and/or unpaid links to web pages, news and/or documents related to theweb page 110 being browsed or the user's activity with any document in their computer to be shown, e.g., infields web page 110 being browsed and the web pages to which ads or links lead to can be assayed by several methods familiar to one skilled in the art, including link analysis, semantic analysis, user behavior etc. Alternatively, a link can be offered to coupons for the web site/profile in question on the toolbar when viewing the corresponding web page. - The present disclosure also includes a toolbar for an Internet browser, similar in aspect to
toolbar 200 discussed above, that provides information on the URL being visited and/or a link to such information. Such information could contain, for example, ads (i.e. advertisings) relevant to the web page being visited. Relevance of ads to a web page can be computed by any measure of semantic similarity between representations of the ad or the web page that clicking on the ad leads to, and of the web page being visited, for example, by latent semantic analysis (LSA) or by the dot product between vectors of word occurrences. Relevance can also be personalized to take into account the probability, based on a user's past behavior (including web pages visited, for example), that the user would be interested in an ad. - A possible field of use of the methods in accordance with the present disclosure is web search. If it is known what pages rank highly for a tag, this can be used to return results during the course of a tag search or a related query. In other words, users can search ‘tagged’ pages on the Web using keywords and filter web pages to be shown only those with a rating above a threshold for a particular tag (e.g. “cool” or “trustworthy”).
- According to a further embodiment, a generic “tagless” rating can be used for generic rankings of web pages and/or people, and used when raters emit a rating (e.g. thumbs up/thumbs down without typing in a tag or description. In some of the embodiments that combine a tagless rating with tags, the raters' tagless or generic rating or reputation is used to weigh their ratings of other people and/or web pages regardless of the tag they employ.
- For some purposes, such as that of assigning reputation to a person, it would be useful to be able to establish whether a web page, such as an online profile of a person, corresponds to a given person. One instance of the problem reduces to that of associating an email address with a web page.
- In accordance with the present disclosure, this is accomplished by providing the person who wishes to verify that they are associated with a given web page with a code (for example a randomly generated string of text and/or numbers and/or other symbols) that the person must insert in the (code for the) corresponding web page. Once the person has placed such string in the web page, the person notifies a service (e.g. clicks on a link) and a server or client verifies whether the code has indeed been inserted in the corresponding web page. In the affirmative case, the association between person and web page gets established. After such event, the person will be free to remove the code from the web page, if it is so desired. The tags and reputation assigned to the web page can subsequently be transferred to the person and affect his/her ratings of other web pages or profiles. In some embodiments, ownership or control of the web page is verified by forcing the owner of the site to place the code in a non-comment section of the web page that is not open for all to modify before the association between web page and user is established.
- The user can be the owner of an email address. Further, the user can be identified by a user identifier, such as a username/password combination or an email address.
- It should also be noted that while the ratings of trustworthy people are better indicators of others' trustworthiness, ratings of beautiful people need not be better indicators of others' beauty. The present disclosure provides a method by which it is possible to automatically detect which tags share a ‘rater to rating transitivity’ (such as the ‘trustworthiness’ rating) and which do not (such as the ‘beauty’ rating). One way of doing that is to measure the correlation between raters' ratings for tag T and the correlation between a rater's rating of web pages or profiles X and the computed score for the corresponding tag ranks of X in a first iteration of the tag rank computation, before the tag rank of a rater influences their influence. If there is a positive correlation, then raters highly rated for that tag are more predictive of the score than raters lowly rated for that tag and should thus be counted more heavily, so the standard tag rank algorithm can be applied. Such tags are said to exhibit rater to rating transitivity (RTRT) and for them, the score computation is performed as described above.
- If there is no statistically significant positive correlation, then no RTRT applies, and the score is computed in the same way as done for tags with negative connotations, or using
Equation 2 above. - In one of the applications of the present disclosure, users of a recommendation network can be allowed to find out how reputable each member in the network is without needing to know how trustworthy each of the recommenders is. This is done by computing, for example, the principal Eigenvector of a recommendation network made up of recommendations or endorsements between individuals, to compute the reputation of each individual.
- According to a further embodiment of the present disclosure, the ratings can also come in the form of communications. For example, if member A sends an e-mail to member B, that could be counted as a vote from A to B. Likewise, if A replies fast to an e-mail from B, that can count as a vote from A to B inversely proportional to (or monotonically decreasing with) the time delay it took A to respond since the e-mail from B was received. Moreover, the use of reputation measured as the eigenvectors in a communication matrix of such communication can be used to combat spam.
- According to a further embodiment of the present disclosure, the reputation of any node N in a network or graph with edges representing positive and negative ‘endorsements’, ratings or reviews can be compared as a function of the corresponding scalar from the dominant eigenvector of the matrix of positive endorsements, the number of negative endorsements received by node N, the number of negative endorsements given by each of the nodes that gave N a negative endorsement, and the reputation of the nodes that gave N a negative endorsement. In some such embodiments, the reputation of node N can be given by
Equation 1 above. - According to another embodiment of the present disclosure, reputation of a user can be visible when the user leaves a comment on a web page of a web site (e.g. a blog). The comment can include computer program code that checks a server for reputation scores and/or tags when a web browser fetches the web page, so that the reputation is constantly updated. Reviews or comments left by the user will be more useful to other users if they know whether that user is trustworthy or not.
- According to still another embodiment, the tag scores can be personalized by counting ratings by people with tastes similar to each user more than others.
- Here is a possible embodiment for a single tag matrix, where a ‘taggable’ subject n is indicated by node n:
-
Personalized TR(node n,tag t,user i)=sum over all users j of [similarity (i,j)×positive— n TR(user j,tag t)×positiverating (node n,tag t,user j)]+sum over all users j of [similarity (i,j)×negative— n TR(user j,tag t)×negativerating (node n,tag t,user j)] (Equation 4) - Where:
-
- rating (node n, tag t, user j) is the rating given by user j to node n; for example, +1 for agree, −1 for disagree and 0 if no rating
- positive_n TR (user j, tag t)=positive-rating-normalized TR=TR (user j, tag t)/sum of all “positive/agree” ratings given by user j for tag t
- negative_n TR (user j, tag t)=negative-rating-normalized TR=TR (user j, tag t)/sum of all “negative/disagree” ratings given by user j for tag t
- And where similarity (i, j) is a measure of the similarity in ratings between user i and user j. For example, one embodiment of similarity is the cosine of the angle between, or the dot product of, the vectors representing users i and j in a matrix which is composed of the N principal components of a matrix M whose rows represent users and whose columns represent nodes (taggable objects) and where each cell contains a user's rating of a taggable object. In other words, a singular value decomposition (SVD) is performed to compute the N principal components of M using principal components analysis (PCA), project each user's row vector into the reduced dimensionality space, and measure their similarity between the projections of two user-vectors as their dot product or the cosine of the angle between them. The cosine is usually preferable as it can be negative and can thus express users with opposite tastes, and it is normalized to the interval between −1 and 1.
- If the units of personalized TR are to be comparable to those of TR, a normalization can be used, so that the sum of all similarities (i, j) over all j equals 1. In other words:
-
Normalized Personalized TR(node n,tag t,user i)=Personalized TR(node n,tag t,user i)/sum over all j of similarity (i,j) (Equation 5) - For multiple interacting tag matrices with predictivities, this generalizes similarly to the way the non-personalized TR does, as shown in the following equation, where PTR stands for personalized TR, R stands for rating (between −1 and +1), and S stands for similarity.
-
PTR(node n,tag t,user i)=Σ(over R)Σ(over j)Σ(over t) of [P(t′→t)·TRj·Sij·Rjnt]/[Σ(over n)Rjt]Σ(over t′) P(t′→t)] (Equation 6) - Equation 6 also includes a normalization, i.e. division by the sum over all ratings by j with tag t. The sum of
Equation 2 includes positive terms in its first portion and negative terms in its second portion. - According to a further embodiment, the trust that user A should have in user B is given also by how much “dilution” has happened between A and B, where dilution is such that each node endorsed by A has 1/n of the trust of A, with n the number of nodes endorsed by A, and each node endorsed by a node X endorsed by A has 1/m of the trust that A has in X, and so on. In other words, the non-global trust can apply to more than just email networks.
- Further embodiments of the present disclosure are also possible, where the teachings of the present disclosure are combined with the teachings of U.S. patent application Ser. No. 11/704,729 filed on Feb. 9, 2007 entitled “Method and System for Authentication of Electronic Communications”, incorporated by reference in the present application.
- While several illustrative embodiments of the invention have been shown and described, numerous variations and alternative embodiments will occur to those skilled in the art. Such variations and alternative embodiments are contemplated, and can be made without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (28)
1. A method for rating an online entity in accordance with a tag, the method comprising:
arranging the online entity to be rated by a plurality of raters in accordance with the tag;
assigning a rater's score to the raters rating the online entity with the tag, to establish reputation of the raters;
assigning a tag score to the online entity, the tag score depending on
i) ratings provided by the raters to the online entity in accordance with the tag, and
ii) the reputation of all raters rating the online entity with the tag.
2-3. (canceled)
4. The method of claim 1 , wherein assigning a tag score to the online entity is established through a raters/entities tag matrix associated with a graph connecting the raters and the online entity.
5-10. (canceled)
11. The method of claim 1 , wherein assigning a rater's score to the raters rating the entity with the tag, to establish reputation of the raters is established by associating each rater with one or more rater's web pages or rater's e-mails and rating the rater's web pages or rater's e-mails in accordance with the tag.
12. The method of claim 11 , wherein when multiple rater's web pages or rater's e-mails are associated with a same rater, the rater's score is calculated by computing initial reputation of each rater's web page and rater's e-mails separately and then providing further reputation of the rater as a function of each initial rater's score.
13. The method of claim 12 , wherein the function is a sum of the initial rater's scores.
14-15. (canceled)
16. A method to establish trust in e-mail communications between senders and receivers, comprising:
assigning a trust score to an online entity represented by an e-mail address, the trust score depending on:
i) amount of communications received by the e-mail address from each sender; and
ii) a trust indication of all senders of the communications to the e-mail address.
17. The method of claim 16 , wherein assigning a trust score to the online entity represented by the e-mail address is established through a senders/online identities matrix associated with a graph connecting the senders and the online identities.
18. The method of claim 16 wherein the trust score further depends on amount of communications sent by the e-mail address.
19. The method of claim 16 , further comprising sending the trust score of the online identity to potential recipients of communications from the online entity.
20. The method of claim 16 , further comprising encrypting the trust score of the online entity.
21. The method of claim 16 , further comprising:
filtering communications receivable from the online entity based on the trust score.
22. The method of claim 21 , wherein the filtering is achieved by selecting a trust score threshold.
23. The method of claim 22 , wherein the trust threshold is variable over time.
24. The method of claim 16 , wherein communications are displayed ranked by the trust score or reputation score of the online entity that sent them.
25. The method of claim 16 , wherein the trust score is further a function of distribution of time delays between a receiver's receipt of communications from the online entity and the receiver's response to the online entity.
26. The method of claim 16 , wherein the trust score is further a function of distribution of length of messages from a receiver to the online entity as compared to length of messages from the online entity to the receiver.
27. The method of claim 16 , wherein the trust score is transferable from a first e-mail address of the online entity to a second e-mail address of the online entity.
28. The method of claim 16 , wherein the trust score is modifiable in accordance with an amount of complaints against the online entity or endorsements in favor of the online entity.
29. The method of claim 16 , wherein the trust score is a function of senders and receivers, such that senders who have been endorsed with communications or ratings from users that are more trusted by a receiver are more trusted than those endorsed by users that are less trusted by the receiver.
30. The method of claim 29 , wherein trust from the receiver radiates outward from the receiver in a network via positive communications or ratings made by the receiver and is diluted with every node away from the receiver.
31-34. (canceled)
35. A method for rating an online entity, the method comprising:
arranging the online entity to be rated by a plurality of raters;
assigning a rater's score to the raters rating the online entity, to establish reputation of the raters; and
assigning a score to the online entity, the score depending on
i) ratings provided by the raters to the online entity, and ii) the reputation of all raters rating the online entity,
wherein assigning the score to the online entity is established through a raters/entities matrix associated with a graph connecting the raters and the online entity.
36-40. (canceled)
41. A method for rating an online entity in accordance with a tag, the method comprising:
providing an agreement identifier, to allow users to agree with the tag;
providing a disagreement identifier, to allow users to disagree with the tag;
assigning a score to the tag in accordance with a difference between agreements with the tag and disagreements with the tag.
42. A method to establish reputation of an online entity, the method comprising:
Establishing a reputation score for the online entity, the reputation score depending on:
i) positive endorsements received by the online entity;
ii) negative endorsements received by the online entity;
iii) negative endorsements given by other entities that gave the online entity a negative endorsement; and
iv) reputation of the entities that gave the online entity a negative endorsement.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US13/196,789 US20120042386A1 (en) | 2006-02-09 | 2011-08-02 | Reputation system for web pages and online entities |
Applications Claiming Priority (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US77206306P | 2006-02-09 | 2006-02-09 | |
US77230106P | 2006-02-09 | 2006-02-09 | |
US11/704,730 US8015484B2 (en) | 2006-02-09 | 2007-02-09 | Reputation system for web pages and online entities |
US13/196,789 US20120042386A1 (en) | 2006-02-09 | 2011-08-02 | Reputation system for web pages and online entities |
Related Parent Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US11/704,730 Continuation US8015484B2 (en) | 2006-02-09 | 2007-02-09 | Reputation system for web pages and online entities |
Publications (1)
Publication Number | Publication Date |
---|---|
US20120042386A1 true US20120042386A1 (en) | 2012-02-16 |
Family
ID=38472505
Family Applications (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US11/704,730 Expired - Fee Related US8015484B2 (en) | 2006-02-09 | 2007-02-09 | Reputation system for web pages and online entities |
US13/196,789 Abandoned US20120042386A1 (en) | 2006-02-09 | 2011-08-02 | Reputation system for web pages and online entities |
Family Applications Before (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US11/704,730 Expired - Fee Related US8015484B2 (en) | 2006-02-09 | 2007-02-09 | Reputation system for web pages and online entities |
Country Status (1)
Country | Link |
---|---|
US (2) | US8015484B2 (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100325107A1 (en) * | 2008-02-22 | 2010-12-23 | Christopher Kenton | Systems and methods for measuring and managing distributed online conversations |
US20110252121A1 (en) * | 2010-04-07 | 2011-10-13 | Microsoft Corporation | Recommendation ranking system with distrust |
US9087109B2 (en) | 2006-04-20 | 2015-07-21 | Veveo, Inc. | User interface methods and systems for selecting and presenting content based on user relationships |
US9177065B1 (en) | 2012-02-09 | 2015-11-03 | Google Inc. | Quality score for posts in social networking services |
US9183259B1 (en) | 2012-01-13 | 2015-11-10 | Google Inc. | Selecting content based on social significance |
US9223835B1 (en) | 2012-01-24 | 2015-12-29 | Google Inc. | Ranking and ordering items in stream |
US9313082B1 (en) | 2011-10-07 | 2016-04-12 | Google Inc. | Promoting user interaction based on user activity in social networking services |
US9319371B1 (en) * | 2011-11-04 | 2016-04-19 | Google Inc. | Management of commercial messages in a social network |
US9454519B1 (en) * | 2012-08-15 | 2016-09-27 | Google Inc. | Promotion and demotion of posts in social networking services |
US20170032471A1 (en) * | 2015-07-30 | 2017-02-02 | Linkedin Corporation | Social proofing for suggested profile edits |
US10811011B2 (en) | 2018-11-21 | 2020-10-20 | Motorola Solutions, Inc. | Correcting for impulse noise in speech recognition systems |
Families Citing this family (131)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8732175B2 (en) | 2005-04-21 | 2014-05-20 | Yahoo! Inc. | Interestingness ranking of media objects |
US10210159B2 (en) * | 2005-04-21 | 2019-02-19 | Oath Inc. | Media object metadata association and ranking |
US8249915B2 (en) * | 2005-08-04 | 2012-08-21 | Iams Anthony L | Computer-implemented method and system for collaborative product evaluation |
US20070078670A1 (en) * | 2005-09-30 | 2007-04-05 | Dave Kushal B | Selecting high quality reviews for display |
US8010480B2 (en) * | 2005-09-30 | 2011-08-30 | Google Inc. | Selecting high quality text within identified reviews for display in review snippets |
US7827052B2 (en) * | 2005-09-30 | 2010-11-02 | Google Inc. | Systems and methods for reputation management |
US8438469B1 (en) | 2005-09-30 | 2013-05-07 | Google Inc. | Embedded review and rating information |
US8549651B2 (en) * | 2007-02-02 | 2013-10-01 | Facebook, Inc. | Determining a trust level in a social network environment |
US9459622B2 (en) | 2007-01-12 | 2016-10-04 | Legalforce, Inc. | Driverless vehicle commerce network and community |
US9070101B2 (en) | 2007-01-12 | 2015-06-30 | Fatdoor, Inc. | Peer-to-peer neighborhood delivery multi-copter and method |
US9071367B2 (en) | 2006-03-17 | 2015-06-30 | Fatdoor, Inc. | Emergency including crime broadcast in a neighborhood social network |
US9373149B2 (en) | 2006-03-17 | 2016-06-21 | Fatdoor, Inc. | Autonomous neighborhood vehicle commerce network and community |
US8965409B2 (en) | 2006-03-17 | 2015-02-24 | Fatdoor, Inc. | User-generated community publication in an online neighborhood social network |
US9098545B2 (en) | 2007-07-10 | 2015-08-04 | Raj Abhyanker | Hot news neighborhood banter in a geo-spatial social network |
US9064288B2 (en) | 2006-03-17 | 2015-06-23 | Fatdoor, Inc. | Government structures and neighborhood leads in a geo-spatial environment |
US9002754B2 (en) | 2006-03-17 | 2015-04-07 | Fatdoor, Inc. | Campaign in a geo-spatial environment |
US9037516B2 (en) | 2006-03-17 | 2015-05-19 | Fatdoor, Inc. | Direct mailing in a geo-spatial environment |
US20080161944A1 (en) * | 2006-12-27 | 2008-07-03 | Lutnick Howard W | Method and apparatus for group filtered reports |
US20080016071A1 (en) * | 2006-07-14 | 2008-01-17 | Bea Systems, Inc. | Using Connections Between Users, Tags and Documents to Rank Documents in an Enterprise Search System |
US7873641B2 (en) | 2006-07-14 | 2011-01-18 | Bea Systems, Inc. | Using tags in an enterprise search system |
US20080016072A1 (en) * | 2006-07-14 | 2008-01-17 | Bea Systems, Inc. | Enterprise-Based Tag System |
US20080016052A1 (en) * | 2006-07-14 | 2008-01-17 | Bea Systems, Inc. | Using Connections Between Users and Documents to Rank Documents in an Enterprise Search System |
US20080016061A1 (en) * | 2006-07-14 | 2008-01-17 | Bea Systems, Inc. | Using a Core Data Structure to Calculate Document Ranks |
US20080016053A1 (en) * | 2006-07-14 | 2008-01-17 | Bea Systems, Inc. | Administration Console to Select Rank Factors |
US7584171B2 (en) * | 2006-11-17 | 2009-09-01 | Yahoo! Inc. | Collaborative-filtering content model for recommending items |
US8843385B2 (en) * | 2006-12-11 | 2014-09-23 | Ecole Polytechnique Federale De Lausanne (Epfl) | Quality of service monitoring of a service level agreement using a client based reputation mechanism encouraging truthful feedback |
US20080147500A1 (en) * | 2006-12-15 | 2008-06-19 | Malcolm Slaney | Serving advertisements using entertainment ratings in a collaborative-filtering system |
US8244599B2 (en) * | 2007-02-28 | 2012-08-14 | Ebay Inc. | Methods and systems for social shopping on a network-based marketplace |
US8346763B2 (en) * | 2007-03-30 | 2013-01-01 | Microsoft Corporation | Ranking method using hyperlinks in blogs |
US7966553B2 (en) * | 2007-06-07 | 2011-06-21 | Microsoft Corporation | Accessible content reputation lookup |
US8635222B2 (en) * | 2007-08-28 | 2014-01-21 | International Business Machines Corporation | Managing user ratings in a web services environment |
US8255975B2 (en) * | 2007-09-05 | 2012-08-28 | Intel Corporation | Method and apparatus for a community-based trust |
US8631015B2 (en) * | 2007-09-06 | 2014-01-14 | Linkedin Corporation | Detecting associates |
US20090077081A1 (en) * | 2007-09-19 | 2009-03-19 | Joydeep Sen Sarma | Attribute-Based Item Similarity Using Collaborative Filtering Techniques |
US20090077093A1 (en) * | 2007-09-19 | 2009-03-19 | Joydeep Sen Sarma | Feature Discretization and Cardinality Reduction Using Collaborative Filtering Techniques |
US7991841B2 (en) * | 2007-10-24 | 2011-08-02 | Microsoft Corporation | Trust-based recommendation systems |
US9367823B1 (en) | 2007-11-09 | 2016-06-14 | Skyword, Inc. | Computer method and system for ranking users in a network community of users |
US8250639B2 (en) | 2007-11-20 | 2012-08-21 | Intel Corporation | Micro and macro trust in a decentralized environment |
WO2009073664A2 (en) * | 2007-12-04 | 2009-06-11 | Google Inc. | Rating raters |
US8291492B2 (en) * | 2007-12-12 | 2012-10-16 | Google Inc. | Authentication of a contributor of online content |
US7840548B2 (en) * | 2007-12-27 | 2010-11-23 | Yahoo! Inc. | System and method for adding identity to web rank |
US7853583B2 (en) * | 2007-12-27 | 2010-12-14 | Yahoo! Inc. | System and method for generating expertise based search results |
US20090192848A1 (en) * | 2008-01-30 | 2009-07-30 | Gerald Rea | Method and apparatus for workforce assessment |
US10055698B2 (en) * | 2008-02-11 | 2018-08-21 | Clearshift Corporation | Online work management system with job division support |
US8442984B1 (en) * | 2008-03-31 | 2013-05-14 | Google Inc. | Website quality signal generation |
US10242104B2 (en) * | 2008-03-31 | 2019-03-26 | Peekanalytics, Inc. | Distributed personal information aggregator |
US20090265198A1 (en) * | 2008-04-22 | 2009-10-22 | Plaxo, Inc. | Reputation Evalution Using a contact Information Database |
US20090306996A1 (en) * | 2008-06-05 | 2009-12-10 | Microsoft Corporation | Rating computation on social networks |
US20090328142A1 (en) * | 2008-06-30 | 2009-12-31 | Industrial Technology Research Institute | Systems and Methods for Webpage Verification Using Data-Hiding Technology |
US20100004982A1 (en) * | 2008-07-03 | 2010-01-07 | Microsoft Corporation | Quantifying trust in computing networks |
US8273182B2 (en) * | 2008-07-15 | 2012-09-25 | WLR Enterprises, LLC | Devices and methods for cleaning and drying ice skate blades |
US8782054B2 (en) * | 2008-07-17 | 2014-07-15 | Microsoft Corporation | Method and system for self-adapting classification of user generated content |
US9002333B1 (en) * | 2008-08-11 | 2015-04-07 | Trend Micro Inc. | Mobile device reputation system |
US9495538B2 (en) * | 2008-09-25 | 2016-11-15 | Symantec Corporation | Graduated enforcement of restrictions according to an application's reputation |
US8364718B2 (en) * | 2008-10-31 | 2013-01-29 | International Business Machines Corporation | Collaborative bookmarking |
US8244664B2 (en) * | 2008-12-01 | 2012-08-14 | Topsy Labs, Inc. | Estimating influence of subjects based on a subject graph |
EP2359276A4 (en) | 2008-12-01 | 2013-01-23 | Topsy Labs Inc | Ranking and selecting enitities based on calculated reputation or influence scores |
US8768759B2 (en) | 2008-12-01 | 2014-07-01 | Topsy Labs, Inc. | Advertising based on influence |
US20100185509A1 (en) * | 2009-01-21 | 2010-07-22 | Yahoo! Inc. | Interest-based ranking system for targeted marketing |
US8312276B2 (en) * | 2009-02-06 | 2012-11-13 | Industrial Technology Research Institute | Method for sending and receiving an evaluation of reputation in a social network |
US8402055B2 (en) * | 2009-03-12 | 2013-03-19 | Desire 2 Learn Incorporated | Systems and methods for providing social electronic learning |
US8312543B1 (en) * | 2009-06-30 | 2012-11-13 | Symantec Corporation | Using URL reputation data to selectively block cookies |
US8661050B2 (en) * | 2009-07-10 | 2014-02-25 | Microsoft Corporation | Hybrid recommendation system |
US8776168B1 (en) | 2009-10-29 | 2014-07-08 | Symantec Corporation | Applying security policy based on behaviorally-derived user risk profiles |
US9110979B2 (en) | 2009-12-01 | 2015-08-18 | Apple Inc. | Search of sources and targets based on relative expertise of the sources |
US11113299B2 (en) | 2009-12-01 | 2021-09-07 | Apple Inc. | System and method for metadata transfer among search entities |
US9129017B2 (en) | 2009-12-01 | 2015-09-08 | Apple Inc. | System and method for metadata transfer among search entities |
US11122009B2 (en) | 2009-12-01 | 2021-09-14 | Apple Inc. | Systems and methods for identifying geographic locations of social media content collected over social networks |
US9454586B2 (en) | 2009-12-01 | 2016-09-27 | Apple Inc. | System and method for customizing analytics based on users media affiliation status |
US9280597B2 (en) | 2009-12-01 | 2016-03-08 | Apple Inc. | System and method for customizing search results from user's perspective |
US8892541B2 (en) | 2009-12-01 | 2014-11-18 | Topsy Labs, Inc. | System and method for query temporality analysis |
US11036810B2 (en) | 2009-12-01 | 2021-06-15 | Apple Inc. | System and method for determining quality of cited objects in search results based on the influence of citing subjects |
US8862699B2 (en) * | 2009-12-14 | 2014-10-14 | Microsoft Corporation | Reputation based redirection service |
KR101123850B1 (en) * | 2010-01-05 | 2012-03-20 | 주식회사 오웨이브미디어 | Method for scoring individual network competitiveness and network effect in an online social network |
US8719352B2 (en) * | 2010-01-29 | 2014-05-06 | Mcafee, Inc. | Reputation management for network content classification |
US8606792B1 (en) | 2010-02-08 | 2013-12-10 | Google Inc. | Scoring authors of posts |
US9953083B2 (en) * | 2010-02-16 | 2018-04-24 | Excalibur Ip, Llc | System and method for determining an authority rank for real time searching |
US8560490B2 (en) * | 2010-02-22 | 2013-10-15 | International Business Machines Corporation | Collaborative networking with optimized inter-domain information quality assessment |
US8527447B2 (en) * | 2010-02-22 | 2013-09-03 | International Business Machines Corporation | Collaborative networking with optimized information quality assessment |
US20110209207A1 (en) * | 2010-02-25 | 2011-08-25 | Oto Technologies, Llc | System and method for generating a threat assessment |
WO2011127206A2 (en) * | 2010-04-06 | 2011-10-13 | Right90, Inc. | Trust rating metric for future event prediction of an outcome |
US9336379B2 (en) * | 2010-08-19 | 2016-05-10 | Microsoft Technology Licensing, Llc | Reputation-based safe access user experience |
WO2012098429A1 (en) * | 2011-01-18 | 2012-07-26 | Nokia Corporation | Method, apparatus, and computer program product for managing unwanted traffic in a wireless network |
GB2502736A (en) | 2011-02-23 | 2013-12-04 | Bottlenose Inc | System and method for analyzing messages in a network or across networks |
GB2488373A (en) * | 2011-02-28 | 2012-08-29 | Hsbc Holdings Plc | Database ranks results based on reputational scores |
US9202200B2 (en) * | 2011-04-27 | 2015-12-01 | Credibility Corp. | Indices for credibility trending, monitoring, and lead generation |
US20120304072A1 (en) * | 2011-05-23 | 2012-11-29 | Microsoft Corporation | Sentiment-based content aggregation and presentation |
US10237060B2 (en) * | 2011-06-23 | 2019-03-19 | Microsoft Technology Licensing, Llc | Media agnostic, distributed, and defendable data retention |
US9189797B2 (en) | 2011-10-26 | 2015-11-17 | Apple Inc. | Systems and methods for sentiment detection, measurement, and normalization over social networks |
US8745738B2 (en) * | 2012-01-15 | 2014-06-03 | Microsoft Corporation | Vouching for user account using social networking relationship |
US8832092B2 (en) | 2012-02-17 | 2014-09-09 | Bottlenose, Inc. | Natural language processing optimized for micro content |
US20130282427A1 (en) * | 2012-04-20 | 2013-10-24 | Martin Dvorak | Reputation-based service valuation |
US9495519B2 (en) * | 2012-05-10 | 2016-11-15 | International Business Machines Corporation | Communications security management |
CN103455515B (en) * | 2012-06-01 | 2017-03-22 | 腾讯科技(深圳)有限公司 | User recommendation method and system in SNS (social networking services) community |
US10019765B2 (en) | 2012-06-27 | 2018-07-10 | Facebook, Inc. | Determining and providing feedback about communications from an application on a social networking platform |
US9009126B2 (en) | 2012-07-31 | 2015-04-14 | Bottlenose, Inc. | Discovering and ranking trending links about topics |
US10164929B2 (en) | 2012-09-28 | 2018-12-25 | Avaya Inc. | Intelligent notification of requests for real-time online interaction via real-time communications and/or markup protocols, and related methods, systems, and computer-readable media |
US9363133B2 (en) | 2012-09-28 | 2016-06-07 | Avaya Inc. | Distributed application of enterprise policies to Web Real-Time Communications (WebRTC) interactive sessions, and related methods, systems, and computer-readable media |
US9411856B1 (en) * | 2012-10-01 | 2016-08-09 | Google Inc. | Overlay generation for sharing a website |
US9047327B2 (en) | 2012-12-03 | 2015-06-02 | Google Technology Holdings LLC | Method and apparatus for developing a social hierarchy |
US9405840B2 (en) * | 2012-12-28 | 2016-08-02 | Microsoft Technology Licensing, Llc | Using social signals to rank search results |
US8762302B1 (en) | 2013-02-22 | 2014-06-24 | Bottlenose, Inc. | System and method for revealing correlations between data streams |
US9294458B2 (en) | 2013-03-14 | 2016-03-22 | Avaya Inc. | Managing identity provider (IdP) identifiers for web real-time communications (WebRTC) interactive flows, and related methods, systems, and computer-readable media |
US10205624B2 (en) | 2013-06-07 | 2019-02-12 | Avaya Inc. | Bandwidth-efficient archiving of real-time interactive flows, and related methods, systems, and computer-readable media |
US9525718B2 (en) | 2013-06-30 | 2016-12-20 | Avaya Inc. | Back-to-back virtual web real-time communications (WebRTC) agents, and related methods, systems, and computer-readable media |
US9065969B2 (en) | 2013-06-30 | 2015-06-23 | Avaya Inc. | Scalable web real-time communications (WebRTC) media engines, and related methods, systems, and computer-readable media |
US9112840B2 (en) | 2013-07-17 | 2015-08-18 | Avaya Inc. | Verifying privacy of web real-time communications (WebRTC) media channels via corresponding WebRTC data channels, and related methods, systems, and computer-readable media |
US9614890B2 (en) | 2013-07-31 | 2017-04-04 | Avaya Inc. | Acquiring and correlating web real-time communications (WEBRTC) interactive flow characteristics, and related methods, systems, and computer-readable media |
US9531808B2 (en) | 2013-08-22 | 2016-12-27 | Avaya Inc. | Providing data resource services within enterprise systems for resource level sharing among multiple applications, and related methods, systems, and computer-readable media |
US10225212B2 (en) | 2013-09-26 | 2019-03-05 | Avaya Inc. | Providing network management based on monitoring quality of service (QOS) characteristics of web real-time communications (WEBRTC) interactive flows, and related methods, systems, and computer-readable media |
US10263952B2 (en) | 2013-10-31 | 2019-04-16 | Avaya Inc. | Providing origin insight for web applications via session traversal utilities for network address translation (STUN) messages, and related methods, systems, and computer-readable media |
US9769214B2 (en) | 2013-11-05 | 2017-09-19 | Avaya Inc. | Providing reliable session initiation protocol (SIP) signaling for web real-time communications (WEBRTC) interactive flows, and related methods, systems, and computer-readable media |
US10129243B2 (en) | 2013-12-27 | 2018-11-13 | Avaya Inc. | Controlling access to traversal using relays around network address translation (TURN) servers using trusted single-use credentials |
CA2880658A1 (en) * | 2014-01-30 | 2015-07-30 | Prabaharan Sivashanmugam | Adaptive social media scoring model with reviewer influence alignment |
US9439367B2 (en) | 2014-02-07 | 2016-09-13 | Arthi Abhyanker | Network enabled gardening with a remotely controllable positioning extension |
US9749363B2 (en) | 2014-04-17 | 2017-08-29 | Avaya Inc. | Application of enterprise policies to web real-time communications (WebRTC) interactive sessions using an enterprise session initiation protocol (SIP) engine, and related methods, systems, and computer-readable media |
US10581927B2 (en) | 2014-04-17 | 2020-03-03 | Avaya Inc. | Providing web real-time communications (WebRTC) media services via WebRTC-enabled media servers, and related methods, systems, and computer-readable media |
US9457901B2 (en) | 2014-04-22 | 2016-10-04 | Fatdoor, Inc. | Quadcopter with a printable payload extension system and method |
US9004396B1 (en) | 2014-04-24 | 2015-04-14 | Fatdoor, Inc. | Skyteboard quadcopter and method |
US9022324B1 (en) | 2014-05-05 | 2015-05-05 | Fatdoor, Inc. | Coordination of aerial vehicles through a central server |
US9441981B2 (en) | 2014-06-20 | 2016-09-13 | Fatdoor, Inc. | Variable bus stops across a bus route in a regional transportation network |
US9971985B2 (en) | 2014-06-20 | 2018-05-15 | Raj Abhyanker | Train based community |
US9912705B2 (en) | 2014-06-24 | 2018-03-06 | Avaya Inc. | Enhancing media characteristics during web real-time communications (WebRTC) interactive sessions by using session initiation protocol (SIP) endpoints, and related methods, systems, and computer-readable media |
US9451020B2 (en) | 2014-07-18 | 2016-09-20 | Legalforce, Inc. | Distributed communication of independent autonomous vehicles to provide redundancy and performance |
CN104519141B (en) * | 2015-01-12 | 2018-07-20 | 张树人 | Quantitative model and method based on relationship evaluation transmission in social relation network |
US10185480B1 (en) * | 2015-06-15 | 2019-01-22 | Symantec Corporation | Systems and methods for automatically making selections in user interfaces |
US11803918B2 (en) | 2015-07-07 | 2023-10-31 | Oracle International Corporation | System and method for identifying experts on arbitrary topics in an enterprise social network |
US11438321B2 (en) * | 2015-12-19 | 2022-09-06 | Telefonaktiebolaget Lm Ericsson (Publ) | Method and apparatus for trust based authentication in SDN clustering |
RU2634180C1 (en) * | 2016-06-24 | 2017-10-24 | Акционерное общество "Лаборатория Касперского" | System and method for determining spam-containing message by topic of message sent via e-mail |
US20180330325A1 (en) | 2017-05-12 | 2018-11-15 | Zippy Inc. | Method for indicating delivery location and software for same |
US11374914B2 (en) | 2020-06-29 | 2022-06-28 | Capital One Services, Llc | Systems and methods for determining knowledge-based authentication questions |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040122926A1 (en) * | 2002-12-23 | 2004-06-24 | Microsoft Corporation, Redmond, Washington. | Reputation system for web services |
US20050097319A1 (en) * | 2003-11-04 | 2005-05-05 | Yahoo! Inc. | System and method for managing a trusted email datastore |
US6895385B1 (en) * | 2000-06-02 | 2005-05-17 | Open Ratings | Method and system for ascribing a reputation to an entity as a rater of other entities |
US20060009994A1 (en) * | 2004-07-07 | 2006-01-12 | Tad Hogg | System and method for reputation rating |
US20060212703A1 (en) * | 2005-03-18 | 2006-09-21 | Fujitsu Limited | E-mail transfer method and device |
US20060253579A1 (en) * | 2005-05-03 | 2006-11-09 | Dixon Christopher J | Indicating website reputations during an electronic commerce transaction |
US20070078699A1 (en) * | 2005-09-30 | 2007-04-05 | Scott James K | Systems and methods for reputation management |
US20080046511A1 (en) * | 2006-08-15 | 2008-02-21 | Richard Skrenta | System and Method for Conducting an Electronic Message Forum |
US7603350B1 (en) * | 2006-05-09 | 2009-10-13 | Google Inc. | Search result ranking based on trust |
US7752554B2 (en) * | 2006-10-05 | 2010-07-06 | Microsoft Corporation | Bot identification and control |
Family Cites Families (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6327656B2 (en) * | 1996-07-03 | 2001-12-04 | Timestamp.Com, Inc. | Apparatus and method for electronic document certification and verification |
US6640301B1 (en) * | 1999-07-08 | 2003-10-28 | David Way Ng | Third-party e-mail authentication service provider using checksum and unknown pad characters with removal of quotation indents |
US20020046250A1 (en) * | 2000-10-17 | 2002-04-18 | Nick Nassiri | Certified and registered electronic mail system |
US20020059525A1 (en) * | 2000-11-10 | 2002-05-16 | Estes Timothy A. | Authenticating the contents of e-documents |
US7188362B2 (en) * | 2001-03-09 | 2007-03-06 | Pascal Brandys | System and method of user and data verification |
US7523496B2 (en) * | 2001-07-31 | 2009-04-21 | International Business Machines Corporation | Authenticating without opening electronic mail |
AUPS193202A0 (en) * | 2002-04-23 | 2002-05-30 | Pickup, Robert Barkley Mr | A method and system for authorising electronic mail |
US20040024823A1 (en) * | 2002-08-01 | 2004-02-05 | Del Monte Michael George | Email authentication system |
US7698558B2 (en) * | 2003-11-21 | 2010-04-13 | Rpost International Limited | System for, and method of, providing the transmission, receipt and content of an e-mail message |
US7444380B1 (en) * | 2004-07-13 | 2008-10-28 | Marc Diamond | Method and system for dispensing and verification of permissions for delivery of electronic messages |
US8484456B2 (en) * | 2004-12-08 | 2013-07-09 | Alien Camel Pty Ltd. | Trusted electronic messaging system |
-
2007
- 2007-02-09 US US11/704,730 patent/US8015484B2/en not_active Expired - Fee Related
-
2011
- 2011-08-02 US US13/196,789 patent/US20120042386A1/en not_active Abandoned
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6895385B1 (en) * | 2000-06-02 | 2005-05-17 | Open Ratings | Method and system for ascribing a reputation to an entity as a rater of other entities |
US20040122926A1 (en) * | 2002-12-23 | 2004-06-24 | Microsoft Corporation, Redmond, Washington. | Reputation system for web services |
US20050097319A1 (en) * | 2003-11-04 | 2005-05-05 | Yahoo! Inc. | System and method for managing a trusted email datastore |
US20060009994A1 (en) * | 2004-07-07 | 2006-01-12 | Tad Hogg | System and method for reputation rating |
US20060212703A1 (en) * | 2005-03-18 | 2006-09-21 | Fujitsu Limited | E-mail transfer method and device |
US20060253579A1 (en) * | 2005-05-03 | 2006-11-09 | Dixon Christopher J | Indicating website reputations during an electronic commerce transaction |
US20070078699A1 (en) * | 2005-09-30 | 2007-04-05 | Scott James K | Systems and methods for reputation management |
US7603350B1 (en) * | 2006-05-09 | 2009-10-13 | Google Inc. | Search result ranking based on trust |
US20080046511A1 (en) * | 2006-08-15 | 2008-02-21 | Richard Skrenta | System and Method for Conducting an Electronic Message Forum |
US7752554B2 (en) * | 2006-10-05 | 2010-07-06 | Microsoft Corporation | Bot identification and control |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9087109B2 (en) | 2006-04-20 | 2015-07-21 | Veveo, Inc. | User interface methods and systems for selecting and presenting content based on user relationships |
US10146840B2 (en) | 2006-04-20 | 2018-12-04 | Veveo, Inc. | User interface methods and systems for selecting and presenting content based on user relationships |
US20100325107A1 (en) * | 2008-02-22 | 2010-12-23 | Christopher Kenton | Systems and methods for measuring and managing distributed online conversations |
US8983975B2 (en) * | 2008-02-22 | 2015-03-17 | Christopher Kenton | Systems and methods for measuring and managing distributed online conversations |
US20110252121A1 (en) * | 2010-04-07 | 2011-10-13 | Microsoft Corporation | Recommendation ranking system with distrust |
US9152969B2 (en) * | 2010-04-07 | 2015-10-06 | Rovi Technologies Corporation | Recommendation ranking system with distrust |
US9313082B1 (en) | 2011-10-07 | 2016-04-12 | Google Inc. | Promoting user interaction based on user activity in social networking services |
US9319371B1 (en) * | 2011-11-04 | 2016-04-19 | Google Inc. | Management of commercial messages in a social network |
US10135780B1 (en) | 2011-11-04 | 2018-11-20 | Google Llc | Management of commercial messages in a social network |
US10536423B1 (en) | 2011-11-04 | 2020-01-14 | Google Llc | Management of commercial messages in a social network |
US9183259B1 (en) | 2012-01-13 | 2015-11-10 | Google Inc. | Selecting content based on social significance |
US9223835B1 (en) | 2012-01-24 | 2015-12-29 | Google Inc. | Ranking and ordering items in stream |
US10133765B1 (en) | 2012-02-09 | 2018-11-20 | Google Llc | Quality score for posts in social networking services |
US9177065B1 (en) | 2012-02-09 | 2015-11-03 | Google Inc. | Quality score for posts in social networking services |
US9454519B1 (en) * | 2012-08-15 | 2016-09-27 | Google Inc. | Promotion and demotion of posts in social networking services |
US20170032471A1 (en) * | 2015-07-30 | 2017-02-02 | Linkedin Corporation | Social proofing for suggested profile edits |
US10811011B2 (en) | 2018-11-21 | 2020-10-20 | Motorola Solutions, Inc. | Correcting for impulse noise in speech recognition systems |
Also Published As
Publication number | Publication date |
---|---|
US20070208613A1 (en) | 2007-09-06 |
US8015484B2 (en) | 2011-09-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US8015484B2 (en) | Reputation system for web pages and online entities | |
US11886555B2 (en) | Online identity reputation | |
US7797421B1 (en) | Method and system for determining and notifying users of undesirable network content | |
US9882790B2 (en) | Method and apparatus for a recommendation system based on token exchange | |
Amintoosi et al. | A reputation framework for social participatory sensing systems | |
CA2794412C (en) | Social graph that includes web pages outside of a social networking system | |
US7774335B1 (en) | Method and system for determining interest levels of online content navigation paths | |
US20130346502A1 (en) | Markup Language for Incorporating Social Networking System Information by an External Website | |
US20110113096A1 (en) | System and method for monitoring activity of a specified user on internet-based social networks | |
US8346950B1 (en) | Hosted application server | |
US20080005223A1 (en) | Reputation data for entities and data processing | |
US20070174249A1 (en) | Method and system for incorporating trusted metadata in a computing environment | |
US20100042577A1 (en) | Systems and methods for calibrating user ratings | |
US20100132049A1 (en) | Leveraging a social graph from a social network for social context in other systems | |
WO2012018443A1 (en) | Integrated link statistics within an application | |
Kurtan et al. | Assisting humans in privacy management: an agent-based approach | |
Buccafurri et al. | A model to support design and development of multiple-social-network applications | |
Parra-Arnau | Pay-per-tracking: A collaborative masking model for web browsing | |
Dasgupta et al. | Enhanced email spam filtering through combining similarity graphs | |
US10922624B2 (en) | Identifying users of shared devices based on user interactions and identity graph | |
US20110161362A1 (en) | Document access monitoring | |
Morales‐del‐Castillo et al. | Recommending biomedical resources: A fuzzy linguistic approach based on semantic web | |
CN114398553A (en) | Object recommendation method and device, electronic equipment and storage medium | |
Puglisi et al. | On web user tracking of browsing patterns for personalised advertising | |
US8429710B1 (en) | Preventing exposure of private information |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
STCB | Information on status: application discontinuation |
Free format text: ABANDONED -- FAILURE TO RESPOND TO AN OFFICE ACTION |