ANNOTATIONS BASED ON HIERARCHICAL CATEGORIES AND GROUPS

Abstract

Systems and methods for recommending entities to a user are presented. In at least one embodiment, a user is identified as belonging to one or more groups of users, or to a hierarchy of groups. A category of entities, preferred by at least of the hierarchy of identified groups, is identified. The category of entities preferred by the at least one identified group corresponds to the category of a user-preferred entity preferred by the user. An entity from the category of entities is selected; the selected entity is not the user-preferred entity. The selected entity is provided to the user as a recommended entity to the user.

Description

BACKGROUND

More and more, people are interacting with and through online services, including but not limited to social networking sites, search engines, online shopping sites, libraries, entertainment/gaming sites, music and video streaming sites, and the like. All of these online services work at a basic level of functionality with each new (or unidentified) user, yet nearly all of these online services work “better” when a user is identified and has provided information about himself/herself to the service. With specific information about the user, these online services are able to “personalize” their services—i.e., provide services specifically tailored and targeted to the user. However, when an online service makes a personalized recommendation to a user, the quality of the personalized recommendation has a direct correlation to user engagement and user satisfaction with that recommendation.

SUMMARY

The following presents a simplified summary in order to provide a basic understanding of various embodiments described herein. This summary is not an extensive overview, and it is not intended to identify key and/or critical elements or to delineate the scope thereof. The sole purpose of this summary is to present some concepts in a simplified form as a prelude to the more detailed description that follows.

According to aspects of the disclosed subject matter, a method for recommending entities to a user is presented. In at least one embodiment, a user is identified as belonging to one or more groups of users, or to a hierarchy of groups. A category of entities, preferred by at least of the hierarchy of identified groups, is identified. The category of entities preferred by the at least one identified group corresponds to the category of a user-preferred entity preferred by the user. An entity from the category of entities is selected; the selected entity is not the user-preferred entity. The selected entity is provided to the user as a recommended entity to the user.

According to additional aspects of the disclosed subject matter, a method embodied on a computer-readable medium bearing computer-executable instructions is presented. The method is configured to provide recommendations of entities to a user. In response to receiving a search query from a user (or other search-triggering event), a plurality of search results is obtained. A group from a hierarchy of groups, to which the user belongs, is identified and a corresponding category of entities that is preferred by the group is identified. An entity is selected of the category of entities. A search results page is generated in response to the search query, the search results page including a subset of the obtained search results, the selected entity and an annotation associated with the selected entity, the annotation identifying that the identified group has an affinity to the selected entity. The search results page is then provided to the user in response to receiving the search query.

BRIEF DESCRIPTION OF THE DRAWINGS

The foregoing aspects and many of the attendant advantages of the disclosed subject matter will become more readily appreciated as they are better understood by reference to the following description when taken in conjunction with the following drawings, wherein:

FIG. 1 is a diagram illustrating an exemplary networked environment suitable for implementing aspects of the disclosed subject matter;

FIG. 2 is a pictorial diagram of an exemplary browser view showing annotated recommended entities in accordance with aspects of the disclosed subject matter;

FIG. 3 is a flow diagram illustrating an exemplary routine suitable for annotating recommended entities on a search results page;

FIG. 4 is a flow diagram illustrating an exemplary sub-routine for determining annotations between a set of entities and a user;

FIG. 5 is a block diagram visually illustrating the annotation of an entity;

FIGS. 6A and 6B show pictorial diagrams for illustrating annotating entities with regard to an image;

FIG. 7 is a flow diagram illustrating an exemplary routine suitable for annotating entities within an image;

FIG. 8 is a block diagram illustrating exemplary components of a search engine configured to respond to search queries with a search results page including annotated entities;

FIG. 9 is a block diagram illustrating exemplary components of an annotation system suitable for annotating a plurality of entities according to aspects of the disclosed subject matter;

FIG. 10 is a pictorial diagram of illustrative user signals used in generating user information for one of many users in the annotation store from which annotations are derived;

FIG. 11 is a pictorial diagram illustrating an exemplary user interface enabling a user to exercise control of the signals used in generating that portion of the annotation store pertaining to the user;

FIG. 12 is a pictorial diagram for illustrating the promotion of a user to a group, and of an entity to a category in determining annotations for the user; and

FIG. 13 is a flow diagram illustrating an exemplary routine for determining a recommendation of an entity of a category to a specific user.

DETAILED DESCRIPTION

For purposed of clarity, the use of the term “exemplary” in this document should be interpreted as serving as an illustration or example of something, and it should not be interpreted as an ideal and/or leading illustration of that thing.

As used in this document, the term “entity” refers to a concept, a person, or a thing. An entity is a “something” which can be annotated. For example, a user will submit a search query including one or more query terms, and these query terms relate to one or more entities—i.e., the intent of the search query. For example, a search query “Paris, France” relates to a single entity, the capital city in France. Search queries may specify multiple entities. For example, the search query “Paris France Eiffel Tower” may be reduced to two entities: (1) the capital of France and (2) the “Eiffel Tower.” A “recommended entity” refers to an entity that has been recommended (typically through personalization) to the user. In the context of a search engine, a recommended entity may include, but is not limited to, a search result (that references suggested content), a suggested search query, a product, an advertisement, and the like. A recommended entity may also comprise a group (or set) of entities and/or a category or subcategory of a product (e.g., “shirts” or “yellow” shirts). For example, a video streaming service may recommend a collection of videos within a genre to the user, the collection being a single recommended entity.

The term, “annotation,” as used throughout this document, refers to a set of relationships between an entity and a user, i.e., the rationale or basis as to how and/or why an entity relates or is relevant to the user. An annotation is comprised of one or more annotation relationships, each relationship describing a single basis for which the user and entity are related. While annotation relationships typically describe a positive affinity between the user and the entity, an annotation relationship may describe a negative affinity between the user and the entity. “Annotating an entity” identifying and associating an annotation with an entity. To visually indicate that an entity has been annotated, an indicator (typically a user-actionable indicator, such as an icon or a hyperlink) is placed in proximity to the entity through which the user can view/access the annotation for that entity. As an alternative to user-actionable indicators, the entire textual annotation may be placed next to the annotated entity.

According to aspects of the disclosed subject matter, an annotation system is present that is configured to annotate one or more entities with regard to a particular user. The annotation system provides an annotation service in which the annotation service receives a set of one or more entities along with the identity of a user and provides annotations from each of the one or more entities.

Advantageously, the annotation system identifies or determines the annotation for an entity independent of their selection or recommendation by another service. In this sense, then, the annotation system is a pluggable system, capable of working with any number of services. This is, in part, accomplished by the fact that the annotation system maintains its own annotation store and annotation analysis engine. With its annotation store and analysis engine, the annotation service issues an annotation independent of the basis by which a cooperating system identifies or recommends the set of entities. For example, a video streaming service may identify a set of videos that it (the video streaming service) wishes to recommend to the user. In annotating the set of videos (either as a group of entities or individually) the annotation system relies upon the information in the annotation store and analysis engine to identify and/or determine the corresponding annotations.

The annotation store includes information (attributes, categories, preferences, relationships, metadata, etc.) about entities, users, and relationships between the two. In conjunction with the information in the annotation store, the annotation service identifies and/or determines a set of annotation relationships between a given entity and user. According to one embodiment annotation relationships between an entity and a user are determined according to probability density functions that predict the likelihood of relevance between the user and the entity.

Clearly, one of the advantages of annotating entities independent of the service that identifies them for annotation is that the cooperative service does not need to gather, ingest, and maintain the robust information that the independent annotation system keeps and uses in annotating entities. In the example above of the video streaming service, the video streaming service may not have access to the identified user's browsing history, the user's purchase history of videos, the user's social network, or any other number of interesting details regarding the user. However, information gathered from these and other sources may be the best rationale of one or more annotation relationships between the entity and the user. Thus, the video streaming service can focus its efforts on providing video streaming services.

While the annotation system may be implemented as a cooperative, stand-alone system, in accordance with aspects of the disclosed subject matter the annotation system may be incorporated within another service. For example, a search engine may be configured to comprise an annotation system such as will be discussed in regard to FIG. 8. However, even when incorporated into another system, the annotation system determines the annotations for a set of entities independent of the process by which the entities were identified or recommended.

Much of the following discussion is made in regard to responding to a search query with from a computer user. While this is one embodiment in which aspects of the disclosed subject matter may operate, and it should be appreciated that the disclosed subject matter is not so limited. Indeed, there are various conditions that may trigger or initiate a search by a search engine or service. User-initiated search queries are search events. Proximity-based apps, such as an app on the user's mobile device for finding restaurants in the device's immediate vicinity, will trigger a search event that obtains search results for the corresponding computer/device user. Recognition services may also cause a search event. For example, a recognition app running on a user's mobile device may initiate a search event to provide information regarding a location or person as the user takes a picture with the mobile device. Accordingly, while much of the discussion that follows is made in regard to responding to a search query from a computer user, it is just one example of a search-triggering event (“search event”) and should not be viewed as limiting upon the disclosed subject matter.

Turning now to FIG. 1, this figure shows is a diagram illustrating an exemplary networked environment 100 suitable for implementing aspects of the disclosed subject matter. The illustrative environment 100 includes one or more user computers, such as user computers 102-106, connected to a network 108, such as the Internet, a wide area network or WAN, and the like. Also connected to the network 108 is a search engine 110. Those skilled in the art will appreciate that a search engine 110 corresponds to an online service hosted on one or more computers on, or computing systems distributed throughout, the network 108. The search engine 110 receives and responds to search queries submitted over the network 108 from various users, such as the users connected to user computers 102-106. In response to receiving a search query, the search engine 110 obtains search results information related and/or relevant to the received search query (as defined by the terms of search query.) The search results information includes search results, i.e., references (typically in the form of hyperlinks) to relevant/related content available from various target sites (such as target sites 112-114) on the network 108. The search results information may also include other information such as related and/or recommended alternative search queries, data and facts regarding the subject matter of the search query, products and/or services related/relevant to the search query, advertisements, and the like. The search engine 110 generates one or more search results pages responsive to the search query based on the search results information. According to various embodiments of the disclosed subject matter, the search engine 110 includes annotated entities with the generated search results pages.

Also shown in the exemplary networked environment 100 is an annotation system 116 for annotating entities, including personalized entities from a search engine 110. While this annotation system 116 is shown as being a separate service/entity in the networked environment 100, it should be appreciated that this is illustrative only and should not be construed as limiting upon the disclosed subject matter. The process of the annotation system in annotating an entity is described in greater detail below.

As those skilled in the art will appreciate, target sites, such as target sites 112-114, host content that is available and/or accessible to users (via user computers) over the network 108. The search engine 110 will be aware of at least some of the content hosted on the many target sites located throughout the network 108, and will store information regarding the hosted content of the target sites in a content index (620 of FIG. 7). The search engine 110 draws from the content index when obtaining search results information in response to receiving a search query. As shown in FIG. 1, the target sites include, by way of illustration, a news organization 112, and a shopping site 114. Of course, those skilled in the art will appreciate that any number and type of target sites may be connected to the network 108. Moreover, as is known in the art, some search engines are aware of millions of target sites and the content that is hosted by those target sites.

Suitable user computers for operating within the illustrative environment 100 include any number of computing devices that can communicate with the search engine 110 or target sites 112-114 over the network 108. In regard to the search engine 110, communication between the user computers 102-106 and the search engine 110 include both submitting search queries and receiving a response in the form of one or more search results pages from the search engine 110. User computers 102-106 may communicate with the network 108 via wired or wireless communication connections. These user computers 102-106 may comprise, but are not limited to: laptop computers such as user computer 102; desktop computers such as user computer 104; mobile phone devices such as user computer 106; tablet computers (not shown); on-board computing systems such as those found in vehicles (not shown); mini- and/or main-frame computers (not shown); and the like.

Turning now to FIG. 2, this figure is a pictorial diagram of an exemplary browser view 200 showing annotated recommended entities within the browser view in accordance with aspects of the disclosed subject matter. In this example, a user has submitted the search query, “Owen Roe Sharecroppers 2008, and browser view 200 illustrates a portion of a generated search results page that includes search results information, including recommended entities. More particularly, the recommended entities include recommended queries 202-204, a recommended search result 206, and a recommended product group 208. As can be seen, each of these recommended entities is annotated with an indicator through which the user can view the rationale for which each entity may have been recommended to the user. For example, a user-actionable icon 210 is placed next to entity group 208 indicating that the entity has been annotated. Annotation view 212 illustrates exemplary annotation relationships between the user who submitted the query and the recommended entity (i.e., the group of wines.) In regard to the annotation information, according to aspects of the disclosed subject matter, for each annotation there may be one or more annotation relationships identified between the user and the entity.

To better understand the process by which entities within a search results page are annotated, reference is now made to FIG. 3. FIG. 3 is a flow diagram illustrating an exemplary routine 300 suitable for annotating recommended entities on a search results page. Beginning at block 302, the search engine 110 receives a search query from a user. At block 304, the search engine 110 obtains search results information responsive to the search query. The search results information includes search results relevant to the search query, i.e., references (typically hyperlinks) to content stored throughout the network 108. Further, however, the search results information will typically include, without limitation, data related to the search query, images, videos, alternative related search queries, search histories, recommended search queries (such as recommended search queries 202-204), related products, and the like.

At block 306, an annotation system associated with the search engine 110 (or incorporated as a part of the search engine) obtains a set of recommended entities from the search results information that was obtained in response to the search query from the user. Once a set of recommended entities is identified, at block 308 those recommended entities are annotated, i.e., annotation information for each recommended entity is obtained. Obtaining annotation information for the recommended entities is described in regard to FIG. 4.

Turning, then, to FIG. 4, this figure is a flow diagram illustrating an exemplary sub-routine 400 for determining annotations between a set of entities and a user. Beginning at block 402, a looping construct is begun to iterate through the set of entities that are to be annotated, such that the steps of blocks 404-408 are repeated for each entity. At block 404, the annotation system determines a set of annotation relationships according to information in a relationship store between the user and the current entity being processed. These relationships are often of a positive nature but may also be a negative relationship (i.e., “product A is recommended because you don't like product B” or “your friend Joe did not like Restaurant X so we do not recommend it to you”). According to various aspects of the disclosed subject matter, the annotation system may determine/identify any number of relationships between the user and the entity, even more than can be reasonably displayed in any one annotation. In limiting the number of annotation relationships that can be used an effort is made to select those relationships that are the strongest, presumptively the most meaningful to the user. Thus, at block 406 an affinity value is assigned to each of the annotation relationships identified between the user and the entity. At block 408, the top n annotation relationships, as determined according to the affinity value between user and the entity, are selected. In this, n is a configurable number as determined according to implementation details, and may be a static or dynamic value.

At block 410, if there are any remaining entities in the set of entities to be annotated the subroutine 400 selects the next entity and returns to block 402 to process that entity. Alternatively, if all of the entities have been annotated, the subroutine 400 proceeds to block 412 where the annotations corresponding to the set of entities is returned.

Returning again to FIG. 3, once the recommended entities are annotated (i.e., are associated with an annotation describing one or more relationships between the user and the entity), the routine 300 proceeds to block 310. At block 310, the search engine 110 generates a search results page. According to aspects of the disclosed subject matter, as part of generating the search results page, the annotated entities are so indicated by an annotation indicator. In general, the annotation indicator is a user-actionable indication that is configured to display the annotation to the user upon its activation. User-actionable icon 210 (FIG. 2) is a non-limiting example of an annotation indicator. After generating the search results page, at block 312 the generated page is returned to the user (in response to the search query) and the routine 300 terminates.

Regarding the routines of FIGS. 3 and 4, (as well as other routines described below) it should be appreciated that while they are expressed with discrete steps, these steps should be viewed as being logical in nature and may or may not correspond to any actual, discrete steps. Nor should the order that these steps are presented be construed as the only order in which the various steps may be carried out. Those skilled in the art will appreciate that logical steps may be combined together or be comprised of multiple steps. Further, while novel aspects of the disclosed subject matter are expressed in routines or methods, this functionality may also be embodied in computer-readable media. As those skilled in the art will appreciate, computer-readable media can host computer-executable instructions for later retrieval and execution. When executed on a computing device, the computer-executable instructions carry out various steps or methods. Examples of computer-readable media include, but are not limited to: optical storage media such as digital video discs (DVDs) and compact discs (CDs); magnetic storage media including hard disk drives, floppy disks, magnetic tape, and the like; transitory and non-transitory memory such as random access memory (RAM), read-only memory (ROM), memory cards, thumb drives, and the like; cloud storage (i.e., an online storage service); and the like. For purposes of this document, however, computer-readable media expressly excludes carrier waves and propagated signals.

In regard to the process by which the annotation system identifies annotations for entities, FIG. 5 is a block diagram visually illustrating the annotation of an entity. As shown in FIG. 5, the top portion of the diagram shows an annotation system 500 that accepts a user identifier 502 as one input to the annotation system and an entity identifier 504 (“Entity_A ID”) as a second input. It should be appreciated that while a single entity (via its identifier) is input into the annotation system 500, this is illustrative only. As discussed above, in an alternative embodiment a group of entities may be submitted to the annotation system 500 for annotation of the group.

Also shown is an annotation store 506 from which the annotation system 500 obtains information regarding the relationships between the entity (as represented by entity identifier 504) and the user (as represented by the user identifier 502). The annotation system 500 obtains the relationship information by way of an analysis engine 514, which analyzes the information from the annotation store (as well as other sources of information) and determines/identifies the various annotation relationships between the user and an entity. The output of the annotation service 500 is the entity annotation 512.

With reference to the lower portion of the diagram, the annotation service obtains a first set 508 of annotation reasons that describe one or more bases for a relationship between the entity and the user—as described in block 404 of FIG. 4. This initial set of annotation relationships is then scored, as shown in the second set 510 of annotation relationships. Generally speaking, these scores represent the strength of affinity between the entity and the user that the annotation relationship represents. In this example, those scores that are the greatest represent the most affinity, but this is illustrative and should not be viewed as limiting on the disclosed subject matter. The third set 512 of annotation relationships represents a selected subset of the annotation relationships of the earlier sets, with these restricted to the n highest scoring relationships, where n=4 (n being a predetermined threshold value). Of course this is illustrative but shows that not all possible annotation relationships need to be included in an annotation.

Regarding the selection of the best (or highest scoring) annotation relationships, while this illustrative diagram shows that the annotation system 500 is responsible for selecting a subset of the best relationships, in an alternative embodiment the annotation system returns all of the identified relationships, along with the affinity scores, such that the requesting service can make the selection itself.

As suggested above, a search engine 110 may be configured with an annotation system (or annotation component) in annotating recommended entities from among search results information. However, the annotation system is not constrained to operate solely as a component of the search engine and, in many cases, operates as an independent service with regard to other online services. Indeed, according to aspects of the disclosed subject matter the annotation system may be implemented as a “pluggable” system that can work (as an independent system) with any number of other systems or services. Examples of this include, but are limited to: associating the annotation system with a video streaming service in which the annotation system annotates video content that the video streaming search recommends to a user; an on-line book store in annotating recommended titles; a social network site in annotating friend and group recommendations; an app or music marketplace; image annotation as described in conjunction with FIGS. 6A-6B and 7; and the like. Annotations may also be made as changes in the current environment occur. For example, the annotation system may annotate offers from local merchants that are periodically sent to subscribers.

In regard to FIGS. 6A and 6B, these two figures show pictorial diagrams for illustrating annotating entities with regard to an image 600. In this example, we can assume that image 600 represents an image that is taken on a user's mobile phone device (that frequently includes a camera for taking images.) A service on the user's mobile phone device (through which the image is taken) provides so-called augmented reality services. This service determines the location and image view just that the subject matter of the image 600 can be recognized. The various entities within the recognized subject matter are then identified and a portion of the entities are recommended to the user. These recommended entities are then submitted to the annotation system (operating as a pluggable service to the augmented reality service, to annotate the recommended entities. An exemplary routine for annotating an image is discussed in regard to routine 700 of FIG. 7. As shown in FIG. 6B, the recommended entities annotated, the augmented reality service overlays the image 600′ with user actionable icons 602-608. Assuming that a user activated icon 602, an annotation window 610 is presented showing one or more annotation relationships between the user and the entity identified in the image 600.

Turning to FIG. 7, this figure is a flow diagram illustrating an exemplary routine 700 suitable for annotating entities within an image, such as the image 600 of FIGS. 6A and 6B. Beginning at block 702, the pluggable annotation system (i.e., an annotation system that operates autonomously or semi-autonomously from other systems) obtains a set of entities within the image that are to be annotated. At block 704, the sub-routine 400 (FIG. 4) is called to annotate the various entities. Thereafter, at block 706, annotation controls (i.e., user-actionable indicators) are overlaid in the image such that the user can interact and discover one or more rationale as to why the corresponding entity is recommended to the user.

FIG. 8 is a block diagram illustrating exemplary components of a search engine 110 configured with an annotation system (as suggested earlier) to respond to search queries with a search results page that includes annotated entities. The search engine 110 includes a processor 802 and a memory 804. As those skilled in the art will appreciate, the processor 802 executes instructions retrieved from memory 804 in carrying out various aspects of the search service, including annotating recommended entities within the search results information.

The search engine 110 also includes a network communications component 806 through which the search engine sends and receives communications over the network 108. For example, it is through the network communication component 806 that the search engine 110 receives search queries from user computers, such as user computers 102-106, and returns results responsive to the search queries. The search engine 110 further includes a search results retrieval component 808 and a search results page generation component 810. Regarding the search results retrieval component 808, this logical component is responsible for retrieving or obtaining search results information relevant to a user's search query from the content index 814. Once the set of search results information responsive to a search query have been retrieved, an entity recommendation component 812 identifies various entities as recommended entities for the user. These recommendations, as well as other personalization information, are typically based on information in a user profile store 816.

It should be appreciated, of course, that many of these components should be viewed as logical components for carrying out various functions of a suitably configured search engine 110. These logical components may or may not correspond directly to actual components. Moreover, in an actual embodiment, these components may be combined together or broke up across multiple actual components.

Also included as part of the search engine 110 is the annotation system. More particularly, this search engine 110 is configured with an annotation system that includes an annotation component 818 that accepts one or more recommended entities and provides an annotation for that entity (as previously described.) Also included as part of the annotation system of the search engine 110 is an annotation store 506 from which the annotation component 818 obtains/identifies the relationships between an entity and the user. In at least one embodiment, these entities are identified through an entity identification and extraction component 820. This entity identification and extraction component identifies a given set of entities with text, such as a user query. Of course, while shown as part of the annotation system portion of the search engine 110, in one embodiment the entity identification and extraction component may be an external component to the search engine.

While the annotation system of FIG. 8 is shown as being a part of the search engine 110, as already mentioned, in an alternative embodiment the annotation system can be implemented as an autonomous or semi-autonomous system. FIG. 9 is a block diagram illustrating exemplary components of an annotation system 900 suitable for annotating a plurality of entities according to aspects of the disclosed subject matter. This exemplary annotation system includes a processor 902 and a memory 904 implementing similar functionality as described above in regard to FIG. 8. The annotation system 900 further includes a network communication component 906 through which the annotation system communicates with other systems in carrying out its annotation function. Also included is an annotation component 818 that accepts one or more entities and provides an annotation for each entity (as previously described in regard to FIG. 8). Still further included as part of the annotation system 900 is an annotation store 506 from which the analysis engine 514 obtains/identifies the relationships between an entity and the user. The entity identification and extraction component, if not included as part of the annotation component 818, identifies a given set of entities from natural language text.

Regarding the various components identified in FIGS. 8 and 9, while certain components are identified as parts of the various computing systems, it should be appreciated that these components should be viewed as logical components for carrying out various functions of suitable configured search engine 110. These logical components may or may not correspond directly to actual components. Moreover, in an actual embodiment, these components may be combined together or broke up across multiple actual components.

As mentioned above in regard to FIG. 5, the annotation system 500 identifies and/or determines the annotation relationships for a given user/entity pair according to information in the annotation store 506. The information in the annotation store includes attributes, aspects, values, and metadata for entities, and similarly includes information regarding users. Numerous user signals—information/data sources relating to a user—are gathered, analyzed and mined for specific pieces of information, attributes, preferences, and the like of a user, and this is done for many users.

Regarding the gathering of information of the users, FIG. 10 is a pictorial diagram of illustrative user signals 1000 used in generating user information for one of many users in the annotation store from which annotations are derived. For each user represented in the annotation store, a set of one or more user signals 1000 is obtained, analyzed and mined to generate the data for the user in the annotation store 506. While user signals 1000 illustrate various illustrative user signals that could be obtained for a given user, they are illustrative only and should not be construed as a limiting or mandatory set of user signals. User profile stores of other services, such as the user profile store 816 of the search engine 800 could be used as a user signal for many users. Moreover, each set of user signals obtained for each user need not be the same as those obtained for another user. For example, while User_A and User_B may belong to a social media site such that social signals 1002 are obtained for each, User_A may have a mobile phone device that enables the user to download apps, thus app signals 1004 (including any apps that have been downloaded by the user, the location of use of the app, whether the app was purchased or free, and the like) may be gathered for User_A, whereas User_B might not download apps to a mobile device but is an online “gamer” such that games signals 1006 are gathered for User_B. Other user signals may include by way of illustration, but not exclusive or limiting, explicit user preferences (that may originate from a user profile store 816 or other source), the geolocation of the user, music preferences (including which songs have been purchased, ratings associated with songs, and the like), and search history information including previously submitted search queries, locations associated and/or specified in search queries, search results that were visited, and the like.

As mentioned, the user signals 1000 are gathered and fed as input into an analysis and mining process 1008. In this process, the signals are converted into specific values, attributes, categories, and the like and are stored in the annotation store 506 and associated with the specific user. Of course, gathering and analyzing various user signals is typically an ongoing, likely periodic, process for all of the users represented in the annotation store 506 as new information may be obtained and the annotation store 506 is updated.

With the ability to gather so many user signals 1000 for any one user, it is advantageous to enable each user to identify whether any signals should be analyzed (i.e., opt in or out) and, if so, which signals can be used. To this end, FIG. 11 is a pictorial diagram illustrating an exemplary user interface 1100 enabling a user to exercise control of the signals used in generating that portion of the annotation store pertaining to the user. The top portion of the user interface 1100 enables the user to opt in or out of annotation system gathering and utilizing the various user signals 1000 in creating user information that is used by the annotation system 500. Assuming that a user is “opted in”, the lower portion of the exemplary user interface 1100 includes various tabs 1102-1114 each corresponding to a specific user signal. As shown in FIG. 1100, the specific user signal shown (via tab 1106) is in regard to the user's music preferences. Moreover, the user may be given an opportunity to opt in or opt out of the annotation system using that specific user signal (as indicated by option 1120, and specific categories within that signal (as indicated by radio buttons 1122-1126). Opting out of any one signal will have the result that the annotation system will not identify annotation relationships between the user and an entity based on that particular signal.

Information provided by an annotation system can be used to provide recommendations to a user. User/entity annotations are scored and the best scoring annotations could be used as the basis for making recommendations. Indeed, there is a strong correlation between the quality of a personalized recommendation and the quality of user engagement with the recommendation, i.e., as the quality of a personalized recommendation increases, so too does the quality of the user interaction with that recommendation. As online services continue to monetize user engagement with items, providing quality, personalized recommendations is important.

One way in which an online service can boost the quality and, therefore, the user engagement with a personalized recommendation is to take advantage of the cognitive behavior of a typical user's need to belong to a group. In particular, users often identify themselves as belonging to networks or groups. Online social media services provide numerous avenues in which a user can associate with a group or network. Moreover, while users will often associate with networks or groups, online social media services (among other online services) may be able to automatically associate a user with one or more groups and/or networks. Often, but not exclusively, this automatic association is based on user online behaviors, preferences, and activities. For example, a Stanford alumnus may identify himself (or be automatically identified) as belonging to that group of people. Moreover, it can often be determined that members of a certain group will prefer certain items, or attend certain functions, and the like. For example, assume it could be determined that Stanford alumni attend a local sports bar whenever Stanford's football team is playing a game. Thus, for User_A who associates himself with the Stanford alumni group, a reasonable recommendation would be the local sports bar during a Stanford football game with an annotation clearly denoting why the sports bar is recommended (i.e., because Stanford alumni apparently prefer to watch the game on the big-screen TV there.)

Similar concepts (making recommendations based on group associations) can be applied to entities. Indeed, it is sometimes more interesting to note that user's aren't as interested in a particular entity as they are in the entity category. For example, a user may watch a specific “James Bond” movie, such as the movie “Quantum Solace,” not because of any particular aspect of the specific movie, but because it is a “James Bond” movie.

By combining the notion of associating a user with one or more groups of users, and categorizing entities into categories, more annotation relationships can be identified and quality recommendations between a user and one or more entities can be made. More particularly, recommendations can be made to a user of items with which the user has no experience, and these recommendations can have a high likelihood of user engagement. Turning, then, to FIG. 12, this is a pictorial diagram 1200 for illustrating the promotion of a user to a group, and of an entity to a category in determining annotations and/or recommendations for the user. To begin with, we assume that user 1202 is known to prefer a particular wine brand, entity 1204. Thus, when annotating the entity 1204, at least one annotation relationship could be something such as, “You often prefer Owen Roe Sharecroppers 2008 vintage.”

However, in expanding the recommendations that may be presented to the user 1202, the process (as will be described below in regard to FIG. 13) determines that the user is a member of group 1206. Since user 1202 likes the wine represented by entity 1204, at least a portion of the group 1206 likes entity 1204. A sampling of the group members can be made to determine a probabilistic statement that members of group 1206 have an affinity to entity 1204. Correspondingly, it may have already been determined (or could similarly be determined) that group 1206 has an affinity to wines represented by the category 1208 and that entity 1204 is a member of category 1208. Moreover, another wine represented by entity 1212 is also a member of the category 1208. Hence, a probability (an affinity value) that a member of group 1206 will also enjoy entity 1212 can be determined/assumed. In other words, since user 1202 is a member of 1206, there is a probability that user 1202 will also have an affinity to entity 1212. In terms of an annotation the relationship if the user 1202 is recommended the entity 1212, the relationship may be as follows: “Your friends of the local wine taster group really like wines like this from Yakima Valley.” The format of this relationship is: “members of group G, of which you are a member, prefer entities of category C, to which entity E belongs.” As can be seen, as users associate themselves with groups, such as user 1210 associating with group 1206, annotation relationships can be formed between the user and entities preferred by the group where no direct relationship between the user and the entities previously existed. Moreover, recommendations can then be made based on these newly established annotation relationships.

Those skilled in the art will appreciate that a user (and/or an entity) will often belong to more than one group (or be automatically associated with multiple groups) that may, in the aggregate, either strengthen or weaken the probability that the user has an affinity to an entity, such as entity 1204. By sampling members of multiple groups to which a user may belong, a stronger probabilistic statement that a user will have an affinity to an entity may be made.

A hierarchy can be established among groups and categories. For example, while not show, group 1206 may belong to a larger group (i.e., members of the group of Stanford Law School alumni are also members of the group of Stanford alumni). The same holds true for categories of entities (i.e., Canon EOS t2i is a member of digital SLR cameras, which is of the category of digital cameras.) Annotation relationships and corresponding recommendations can be identified by traversing up the hierarchy of groups and entities.

FIG. 13 is a flow diagram illustrating an exemplary routine for determining a recommendation of an entity of a category to a specific user. Beginning with block 1302, a selected group of all of the groups to which a user belongs are identified. This selection/identification of a group may include identification from social groups, super-groups, as well as a group of a hierarchy of groups to which the user belongs, either directly or as a group belonging to another group. Moreover, inclusion within a group may be an explicit inclusion by the user (or another party) or by implicit/implied inclusion of a group.

At block 1304, a category (of potentially many categories) that is preferred by the selected group is identified. As above, this preferences or affinity between the group and the identified category may be based on explicit preference or according to an implicit/implied affinity or preference. At block 1306, affinity values between the entities of the identified category and the selected group are determined. In at least one embodiment, the affinity values are determined according to probability that the user will prefer the corresponding entity using one or more probability density functions. At block 1308, the entity having the greatest affinity value is selected for recommendation to the user. At block 1310, the selected pair is returned. As this is a recommendation process, in at least one embodiment the selected entity is not already a user-preferred entity, such as in the example discussed above in regard entity 1212 (FIG. 12) that was not related to user 1202.

As mentioned in block 1310, the recommended entity is accompanied by an annotation that identifies the basis on which the selected entity is recommended to the user. By way of example, for the returned entity the annotation relationship underlying the annotation may be “Members of Group G, of which you are a member, tend to prefer Entity E.” Thereafter, the routine 1300 terminates.

While routine 1300 is presented as an entity recommendation routine, it should be appreciated that this routine could be included in any number settings, including within a process of responding to a search query. Moreover, in at least one embodiment, the category is selected according to the subject matter of the search query. By way of example and not limitation, when search results obtained in response to a search query do not correspond to an entity with which the user already has a relationship, the search engine can provide a recommendation (and annotation) to the user of an entity that corresponds to an obtained search result. For example (and with reference to FIG. 12), should user 1202 submit a search query for wines that would not include a search result directed to entity/wine 1204, yet would include search results for wines from the category 1208 of wines, then the search engine could recommend entity/wine 1212 that is a member of category 1208.

Throughout the previous discussion, reference has been made to a discover system for identifying relationships between a user and an entity based on hierarchies of groups and classes of entities in the context of a user submitting a search query to a search engine. However, as suggested in regard to FIG. 9, the disclosed subject matter should not be viewed as limited to the search query/search engine context. It is anticipated that any number of environments may be enhanced according to aspects of the disclosed subject matter. By way of example, but not limitation: a streaming video service may classify a user into one or more groups (or allow a user to self-classify) and a hierarchy of groups may be used to recommend one or more videos to the user according to affinities between the groups and genres (classifications) of videos; a college guidance system may obtain information about a student (both explicit and implicit information) and provide highly relevant guidance information to the student regarding colleges (based on classification hierarchies of the various educational institutions); and the like.

While various novel aspects of the disclosed subject matter have been described, it should be appreciated that these aspects are exemplary and should not be construed as limiting. Variations and alterations to the various aspects may be made without departing from the scope of the disclosed subject matter.

Claims

1. A computer-implemented method for providing a recommended entity to a user, the method comprising:

identifying a group to which the user belongs;

identifying a category of entities preferred by the identified group, wherein the category of entities preferred by the identified group corresponds to the category of a user-preferred entity preferred by the user;

selecting an entity of the category of entities having the greatest affinity value between the group and the selected entity, wherein the selected entity is not the user-preferred entity; and

providing the selected entity as a recommended entity to the user and associating an annotation with the selected entity, the annotation including an annotation relationship identifying that the identified group has an affinity to the selected entity.

2. The method of claim 1 further comprising:

obtaining a plurality of search results responsive to a search query received from the user; and

generating a search results page, the search results page including a subset of the obtained plurality of search results and further including the recommended entity and the associated annotation;

wherein providing the selected entity as a recommended entity to the user comprises providing the generated search results page to the user.

3. The method of claim 2, wherein at least one search result of the subset of search results included in the generated search results page corresponds to the selected entity, and wherein the annotation associated with then entity is placed proximately to the at least one search result in the generated search results page.

4. The method of claim 1, wherein the annotation relationship describes a basis for which the recommended entity is relevant to the identified group.

5. The method of claim 1, wherein the annotation relationship describes a positive affinity between the identified group and the recommended entity.

6. The method of claim 1, wherein the annotation relationship describes a negative affinity between the identified group and the recommended entity.

7. The method of claim 1, wherein the affinity values between the group and the entities of the category of entities are determined according to one or more probability density functions.

8. The method of claim 1, wherein the search query is directed to items of the category of entities.

9. A computer-readable medium bearing computer-executable instructions which, when executed on a computing device having at least a processor and a memory, carry out a method for providing recommendations of entities to a user, the method comprising:

obtaining a plurality of search results responsive to receiving a search query from a user;

determining a group from a hierarchy of groups to which the user belongs;

identifying a plurality of entities of a category that are preferred by the group;

for each of the plurality of entities, determining an affinity value between the group and the entity;

selecting the entity having the greatest affinity value to the group;

generating a search results page responsive to the search query, the search results page including: a subset of the obtained search results; the selected entity presented as a recommended entity to the user; and an annotation, the annotation comprising an annotation relationship identifying that the identified group has an affinity to the selected entity; and

providing the generated search results page to the user responsive to the search query.

10. The computer-readable medium of claim 9, wherein at least one search result of the subset of search results included in the generated search results page corresponds to the selected entity, and wherein the annotation associated with then entity is placed proximately to the at least one search result.

11. The computer-readable medium of claim 9, wherein the annotation relationship describes a basis for which the recommended entity is relevant to the identified group.

12. The computer-readable medium of claim 9, wherein the annotation relationship describes a positive affinity between the identified group and the recommended entity.

13. The computer-readable medium of claim 9, wherein the annotation relationship describes a negative affinity between the identified group and the recommended entity.

14. The computer-readable medium of claim 9, wherein the affinity values between the group and the entities of the category of entities are determined according to one or more probability density functions.

15. The computer-readable medium of claim 9, wherein the search query is directed to items of the category of entities.

16. A system for recommending an entity to a user, wherein the recommendation system is implemented on a computer system comprising at least a processor executing instructions stored in a memory, and further comprising:

an annotation store, the annotation store storing relationship information between a plurality of groups of users and a plurality of entities; and

a recommendation component that, responsive to an entity recommendation for a user: identifies a group of users to which the user belongs; identifies a category of entities preferred by the identified group, wherein the category of entities preferred by the identified group corresponds to the category of a user-preferred entity preferred by the user; and selects an entity of the category of entities having the greatest affinity value between the group and the selected entity, wherein the selected entity is not the user-preferred entity;

wherein the system provides the selected entity as a recommended entity to the user.

17. The system of claim 16 further comprising an annotation component that associates an annotation with the selected entity, and wherein the system provides the selected entity as a recommended entity to the user with the associated annotation.

18. The system of claim 17 further comprising a search results retrieval component that obtains a plurality of search results responsive to the system receiving a search query from the user; and

a search results page generator that generates a search results page responsive to the system receiving the search query from the user, the search results page including a subset of the obtained plurality of search results and further including the recommended entity and the associated annotation.

19. The system of claim 18, wherein at least one search result of the subset of search results included in the generated search results page corresponds to the selected entity, and wherein the associated annotation is placed proximately to the at least one search result in the generated search results page.

20. The system of claim 19, wherein the affinity values between the group and the entities of the category of entities are determined according to one or more probability density functions.