Abstract: A system uses semantic analysis of text associated with content items to recommend content for display to a user. A subset of representative words from a content description are determined and a content embedding that models the content is generated using a combination of word embeddings associated with each of the representative words. User embeddings are generated using a combination of content embeddings for content that a user has had particular interactions with in a set period of time. Separate user embeddings may be generated to represent user interactions with different categories of content (e.g., travel, photography, apparel, comedy, etc.). The system uses the content embeddings and user embeddings as input to predictive functions which determine a candidate content item that a user is likely to interact with if the candidate content is displayed to the user.

Abstract: Systems, methods, and non-transitory computer-readable media can receive a set of candidate training items for training an early stage model in a multi-stage recall optimization model, wherein the multi-stage recall optimization model comprises the early stage model and a target model. A random subset of the candidate training items is selected from the set of candidate training items. For each training item in the subset of candidate training items, a score is determined based on the target model. Each training item in the subset of candidate training items is labeled with a label based on a probability of the training item being a top-K of the set of candidate training items had the set of candidate training items been scored based on the target model.

Abstract: An on-line social network system is configured to generate a news feed for a member by processing updates originating from different sources using different first pass ranker models. The first pass ranker models generate respective sets of raw scores, which are calibrated based on a consistent scale of feed engagement metrics of interest, such as a click through rate. The calibrated scores are then used as training data to train a second pass ranker and/or as input into the second pass ranker at the time when the second pass ranker is to generate respective ranks for items in an inventory of updates identified as potentially of interest to a focus member and to select a subset of items from the inventory based on the generated respective ranks.

Abstract: Disclosed in some examples are methods, systems, and machine-readable mediums which automatically determine network-based data sources for information ingestion and profile data completion. This method can be applied to automatically increase the library of network-based data sources utilized by the system to ingest profile information. This allows for more a complete tracking of member accomplishments and attributes and ultimately, allows for more complete member profiles. Before specific methods and systems for automatically determining network-based data sources are discussed, an overview of the process of ingesting information from network-based data sources and matching that information to members of the social networking service will be described.

Abstract: In one embodiment, a method includes receiving a query and determining a query vector. The method includes accessing multiple object vectors representing multiple objects, respectively. The method includes, for a first set of object vectors identified as top object vectors, calculating an inner product with the query vector. The method includes progressively computing an inner product of the query vector and each remaining object vector and sending, to a user, the objects corresponding to the top object vectors. Progressively computing an inner product includes checking whether to calculate a first partial inner product based on a bound on the inner product and the minimum inner product for a top object vector, calculating subsequent partial inner products until the inner product is complete, and substituting the object vector for a top object vector if the complete inner product is greater than the minimum inner product.

Abstract: Techniques for presenting a personalized member profile page to a viewer are described. The online social network service system can access a sender score of a sender requesting to view a profile page in an online social network service, and access a receiver score of a receiver associated with the profile page. Additionally, a suggested action can be determined based on the sender score and the receiver score, a sender confirmation to perform the suggested action can be received, and a communication associated with the suggested action can be transmitted in response to the received sender confirmation. Subsequently, the online social network service system can classify an interaction between the sender and the receiver, and update the receiver score and the sender score based on the classified interaction.

Abstract: An online system selects items for display in content provided to users by considering the value of each item to third-party content providers as well as user's interests. The online system receives a catalog including items that are each associated with weights from a third-party content provider for inclusion in sponsored content to be presented to users of an online system. The weights have values indicating measures of importance of the items to the third-party content provider on a per-item basis. The online system identifies a request for sponsored content, and selects one or more items from the catalog for inclusion in a dynamic sponsored content item. The online system calculates a weighted user preference score using a weight associated with an item and affinity information describing the user's affinity for the item.

Abstract: In one embodiment, a method includes receiving a query and determining a query vector. The method includes accessing multiple object vectors representing multiple objects, respectively. The method includes, for a first set of object vectors identified as top object vectors, calculating an inner product with the query vector. The method includes progressively computing an inner product of the query vector and each remaining object vector and sending, to a user, the objects corresponding to the top object vectors. Progressively computing an inner product includes checking whether to calculate a first partial inner product based on a bound on the inner product and the minimum inner product for a top object vector, calculating subsequent partial inner products until the inner product is complete, and substituting the object vector for a top object vector if the complete inner product is greater than the minimum inner product.

Abstract: A news feed system of an on-line social network system obtains and utilizes data related to events that originate with members of the on-line social network system from web pages other those members' news feed pages. The news feed system monitors signals that are not related to feed interaction and generates contextual engagement features based on those signals. The news feed system associates contextual engagement features with respective member profiles and may store the association information for a period of time. The news feed system then uses these features to train a model that ranks news feed inventory and/or as input into that model.

Abstract: A news feed system of an on-line social network system news utilizes a relevance model to determine which updates from an inventory of updates are to be presented to a member on their news feed page. The relevance model is trained using historical data that reflects interactions of members of the on-line social network system with items in their respective news feed pages. In order to reduce potential biases in the historical data that is used to train the relevance model, the news feed system designates a certain portion of all member sessions to be random sessions. The news feed generated for a member during a random session includes updates that are selected and/or ordered for presentation using one or more randomization techniques.

Abstract: A method and system to suggest keywords to a social network member is described. A suggested keywords system, in one example embodiment, examines phrases that appear in profiles maintained by the on-line social networking system that are similar to the target profile and identifies those words and phrases that are most prominent in these profiles, utilizing a graph-based approach. These most prominent words and phrases may be presented to the target member as suggested keywords to be included in the member's professional summary.

Abstract: A suggested keywords system is configured for identifying phrases, which are most relevant to experience and expertise of a professional network member, and which the member may be interested in weaving into their profile summary. The suggested keywords system generates a model, for each phrase, that calculates probability of that phrase being present in a profile that is characterized by the absence of certain attributes and by the presence of certain attributes. Based on the model, the suggested keywords system calculates a ranking value for the phrase for a particular target profile. The phrases with the higher rank are considered to be more relevant in describing professional background of the target member. A certain number of phrases that have the highest ranking are presented to the member as suggested keywords to be included in their professional summary.

Abstract: Method and system to determine a company rank utilizing on-line social network data are described. A company ranking system is configured to access a base set of member profiles, construct a talent flow graph having nodes that represent respective companies and edges that represent transitions of employees from one company to another, and determine a node score for each node by applying a ranking algorithm to the graph. In one embodiment, a company ranking system generates perturbed versions of the base set by using bootstrap resampling procedure and uses the perturbed versions of the base set to calculate ranking data for the nodes in the talent flow graph in the form of node scores. The distribution of the node scores included in the ranking data calculated for a given node is used to determine a desirability score for the company represented by the node.

Abstract: A method is disclosed for supplementing information that is to be presented to a user of a social-networking system with respect to a job posting. A request for a job posting stored in a database of a job hosting service is received from a client-computing device. The job posting is for a job opening of an employer and the request is associated with a first member of a social networking service. A snippet of a member profile of a second member of the social networking service is selected for presentation with the job posting. The snippet is extracted from the member profile based on an identification of a correspondence between information specified in the member profile and information specified in the job posting. Responsive to the request, the snippet is communicated to the client-computing device for presentation to the first member in conjunction with the job description.

Abstract: Techniques for optimizing a guided edit process for editing a member profile page are described. According to various embodiments, incomplete member profile fields in a member profile associated with member of an online social networking service are identified. Profile completion score weight values associated with the incomplete member profile fields in the member profile are then determined. Thereafter, the incomplete member profile fields in the member profile are ranked, based on the profile completion score weight values associated with each of the incomplete member profile fields. A list of one or more of the highest ranked incomplete member profile fields are then displayed, together with a prompt recommending the member to complete the identified incomplete member profile fields.

Abstract: A relevance model is used to process an inventory of updates for a member of an on-line social network in order to select a subset of updates for presentation to the member. One of the features used as input to the relevance model is viewer affinity. The viewer affinity indicates preference of a member for a particular type or source of information and is determined using the estimated probability of the member clicking on the impression of an update and also based on a correction variable. The correction variable is generated based on information regarding previously-observed interactions of the member with the updates.

Abstract: An on-line social network system is configured to generate a news feed for a member by processing updates originating from different sources using different first pass ranker models. The first pass ranker models generate respective sets of raw scores, which are calibrated based on a consistent scale of feed engagement metrics of interest, such as a click through rate. The calibrated scores are then used as training data to train a second pass ranker and/or as input into the second pass ranker at the time when the second pass ranker is to generate respective ranks for items in an inventory of updates identified as potentially of interest to a focus member and to select a subset of items from the inventory based on the generated respective ranks.

Abstract: A news feed system of an on-line social network system obtains and utilizes data related to events that originate with members of the on-line social network system from web pages other those members' news feed pages. The news feed system monitors signals that are not related to feed interaction and generates contextual engagement features based on those signals. The news feed system associates contextual engagement features with respective member profiles and may store the association information for a period of time. The news feed system then uses these features to train a model that ranks news feed inventory and/or as input into that model.

Abstract: An on-line social network system includes a ranker to processes an inventory of news feed updates for a member and select more relevant updates for presentation to the member. The ranker is trained using training data that includes personalized engagement probability for an update. The personalized engagement probability values are calculated in real time and for a particular update with respect to member features that appear in member profiles maintained by the on-line social network system.

Abstract: Techniques for presenting a personalized member profile page to a viewer are described. A highlight module can receive a request to view a profile page of a member in a social network. The highlight module can access viewer data of a viewer associated with the request, and access member data of the member. Additionally, the highlight module can determine a plurality of member attributes relevant to the viewer based on the viewer data, the plurality of member attributes being derived from the member data. Furthermore, the highlight module can calculate an overall score for a member attribute in the plurality of member attributes based on the viewer data and the member data. Subsequently, a profile generation module can cause a presentation, on a display of a device, of the member attribute on the profile page, when the overall score of the member attribute is higher than a predetermined threshold value.