Abstract: Techniques for implementing a learning semantic representations of sparse entities using unsupervised embeddings are disclosed herein. In some embodiments, a computer system accesses corresponding profile data of users indicating at least one entity of a first facet type associated with the user, and generating a graph data structure comprising nodes and edges based on the accessed profile data, with each node corresponding to a different entity indicated by the accessed profile data, and each edge directly connecting a different pair of nodes and indicating a number of users whose profile data indicates both entities of the pair of nodes. The computer system generating a corresponding embedding vector for the entities based on the graph data structure using an unsupervised machine learning algorithm.

Abstract: Techniques for query intent clustering for automated sourcing are described. In an example embodiment, disclosed is a system comprising a processor, a storage device, and a memory device holding an instruction set executable on the processor to cause the system to perform operations. The system obtains one or more recent hire member profiles used as a basis for a search on member profiles in a social networking service. Additionally, the system extracts one or more attributes from the one or more recent hire member profiles and stores the attributes on the storage device. Moreover, the system identifies skills clusters based on the extracted attributes retrieved from the storage device. Furthermore, the system generates a search query based on the identified skills clusters. Then, a search can be performed on member profiles in the social networking service using the generated search query, returning one or more result member profiles as candidates.

Abstract: In an example embodiment, gradient boosted decision trees are used to generate tree interaction features, which encode a set of decision rules for features of search results and hence allow feature interactions. These tree interaction features may then be used as features of a GLMix model, essentially injecting non-linearity into the GLMix model.

Abstract: Techniques for providing explanations of candidate search queries are described. The queries can be created using query intent clustering in an automated sourcing tool. In an example embodiment, disclosed is a system that obtains one or more current candidate member profiles used as a basis for a search on member profiles in an online system. Additionally, the system extracts one or more attributes from the one or more current candidate member profiles. Moreover, the system identifies query intent clusters based on the extracted one or more attributes. Furthermore, the system generates a search query based on the identified query intent clusters. Next, an explanation of the search query can be displayed on a display device of the system. In some embodiments, the online system hosts a social networking service that includes the member profiles, and the identified query intent clusters include skills clusters.

Abstract: The disclosed embodiments provide a system for performing dynamic job bidding optimization. During operation, the system obtains historical data containing a time series of interactions with a job. Next, the system uses the historical data to calculate an initial price of a job based on a predicted number of interactions with the job. The system then determines a first dynamic adjustment to the initial price that improves utilization of a budget for the job and a second dynamic adjustment to the initial price that improves a performance of the job. Finally, the system applies the first and second adjustments to the initial price to produce an updated price for the job and delivers the job within an online system based on the updated price.

Abstract: Techniques for processing search queries are described. Consistent with some embodiments, a computer system generates a profile vector representation for each of several user profiles based on the user profile data of the user profiles, and then stores the vector representations. A subsequent query is processed to generate a query vector representation for the query. A neural network is used to generate a similarity score for each pairing of the query vector representation and a profile vector representation. Finally, some number of user profiles having the highest similarity scores are provided as search results.

Abstract: Techniques for using recruiter review data to create training, validation and test sets for automated sourcing are described. An example system obtains sample suggested candidate member profiles and sample search result member profiles in an online system. The system identifies unique pairs of member profiles, each pair consisting of one of the suggested candidate profiles and one of the search result profiles. Additionally, the system generates a label for each of the unique pairs of profiles. The label is generated using a score generated from log information of the online system, the log information including historical records of searcher feedback regarding members of the online system, the score being higher if the searcher accepted the sample search result member profile in a search session. Furthermore, the system inputs the labels into a machine learning algorithm to train a combined ranking model that outputs ranking scores for search result member profiles.

Abstract: In an example embodiment, gradient boosted decision trees are used to generate tree interaction features, which encode a set of decision rules for features of search results and hence allow feature interactions. These tree interaction features may then be used as features of a GLMix model, essentially injecting non-linearity into the GLMix model.

Abstract: Techniques for applying learning-to-rank with deep learning models for search are disclosed herein. In some embodiments, a computer system trains a ranking model using training data and a loss function, with the ranking model comprising a deep learning model and being configured to generate similarity scores based on a determined level of similarity between profile data of reference candidates users in the training data and reference query data of reference queries in the training data. The computer system receives a target query comprising target query data from a computing device of a target querying user, and then generates a corresponding score for target candidate users based on a determined level of similarity between profile data of the target candidate users and the target query data using the trained ranking model.

Abstract: Disclosed are systems, methods, and non-transitory computer-readable media for generating personalized insights. An insight generation system, in response to a first user of an online service having added a second user as a recipient of a message, gathers profile data of the second user and profile data of an entity that is maintained by the online service. The insight generation system determines a set of insights for the second user, based on the profile data of the second user, the profile data of the entity, and a set of insight algorithms. Each insight indicates commonalities between the second user and the entity. The insight generation system selects a subset of the set of insights, yielding a set of recommended insights, and provides the set of recommended insights to a client device of the first user.

Abstract: Techniques for generating supervised embedding representations for search are disclosed herein. In some embodiments, a computer system receives training data comprising query representations including an entity included in a corresponding search query submitted by a querying user, search result representations for each one of the query representations, and user actions for each one of the query representations, and generates a corresponding embedding vector for each one of the at least one entity using a supervised learning algorithm and the received training data. In some example embodiments, the corresponding search result representations for each one of the query representations represents a plurality of candidate users displayed in response to the search queries based on profile data of the candidate users, and the user actions comprise actions by the querying user directed towards at least one candidate user in the search results.

Abstract: Techniques for generating candidates for search using a scoring and retrieval architecture and deep semantic features are disclosed herein. In some embodiments, a computer system generates a profile vector representation for user profiles based profile data, stores the profile vector representations, receives a query subsequent to the storing of the profile vector representations, generates a query vector representation for the query, retrieves the stored profile vector representations of the user profiles based on the receiving of the query, generates a corresponding score for pairings of the user profiles and the query based on a determined level of similarity between the profile vector representation of the user profiles and the query vector representation, and causes an indication of at least a portion of the user profiles to be displayed as search results for the query based on the generated scores of the user profiles.

Abstract: Techniques for implementing a learning semantic representations of sparse entities using unsupervised embeddings are disclosed herein. In some embodiments, a computer system accesses corresponding profile data of users indicating at least one entity of a first facet type associated with the user, and generating a graph data structure comprising nodes and edges based on the accessed profile data, with each node corresponding to a different entity indicated by the accessed profile data, and each edge directly connecting a different pair of nodes and indicating a number of users whose profile data indicates both entities of the pair of nodes. The computer system generating a corresponding embedding vector for the entities based on the graph data structure using an unsupervised machine learning algorithm.

Abstract: Techniques for generating supervised embedding representations using unsupervised embedding representations and deep semantic structured models for search are disclosed herein. In some embodiments, a computer system generates a graph data structure based on accessed profile data, generates an initial embedding vector using an unsupervised machine learning algorithm, receiving training data comprising query representations, search result representations, and user actions, with each one of the plurality of query representations comprising the initial embedding vector, and generates a final embedding vector using a supervised learning algorithm and the received training data.

Abstract: Systems and methods for query expansion are disclosed. In some examples, a server receives, from a client device, a search query for employment candidates, the search query comprising a first set of parameters. The server determines a second set of parameters related to the first set of parameters in response to identifying a second parameter for the second set of parameters that corresponds with a first parameter from the first set of parameters, the professional records being stored in a professional data repository. The server generates, from the professional data repository, a first set of search results based on the first set of parameters and the second set of parameters. The server provides, to the client device, an output representing the first set of search results.

Abstract: Techniques for providing explanations of candidate search queries are described. The queries can be created using query intent clustering in an automated sourcing tool. In an example embodiment, disclosed is a system that obtains one or more current candidate member profiles used as a basis for a search on member profiles in an online system. Additionally, the system extracts one or more attributes from the one or more current candidate member profiles. Moreover, the system identifies query intent clusters based on the extracted one or more attributes. Furthermore, the system generates a search query based on the identified query intent clusters. Next, an explanation of the search query can be displayed on a display device of the system. In some embodiments, the online system hosts a social networking service that includes the member profiles, and the identified query intent clusters include skills clusters.

Abstract: Techniques for using recruiter review data to create training, validation and test sets for automated sourcing are described. An example system obtains sample suggested candidate member profiles and sample search result member profiles in an online system. The system identifies unique pairs of member profiles, each pair consisting of one of the suggested candidate profiles and one of the search result profiles. Additionally, the system generates a label for each of the unique pairs of profiles. The label is generated using a score generated from log information of the online system, the log information including historical records of searcher feedback regarding members of the online system, the score being higher if the searcher accepted the sample search result member profile in a search session. Furthermore, the system inputs the labels into a machine learning algorithm to train a combined ranking model that outputs ranking scores for search result member profiles.

Abstract: Techniques for dynamically altering weights to re-weight candidate features of a candidate search and ranking model in a streaming environment are described. In an embodiment, a disclosed system obtains desired hire documents using a search query specifying parameters. Additionally, the system extracts desired hire-based features from the documents, with the features corresponding to the parameters. Moreover, the system inputs the features to a combined ranking model that is trained by a machine learning algorithm to output a ranking score for each of the documents, with the combined ranking model including weights assigned to each of the features. Furthermore, the system ranks the desired hire documents based on the ranking scores and displays top ranked documents. Then, feedback is received regarding the top ranked documents, and the weights assigned to each of the features are dynamically trained to alter the weights assigned to each of the features based on the feedback.

Abstract: Techniques for query intent clustering for automated sourcing are described. In an example embodiment, disclosed is a system comprising a processor, a storage device, and a memory device holding an instruction set executable on the processor to cause the system to perform operations. The system obtains one or more recent hire member profiles used as a basis for a search on member profiles in a social networking service. Additionally, the system extracts one or more attributes from the one or more recent hire member profiles and stores the attributes on the storage device. Moreover, the system identifies skills clusters based on the extracted attributes retrieved from the storage device. Furthermore, the system generates a search query based on the identified skills clusters. Then, a search can be performed on member profiles in the social networking service using the generated search query, returning one or more result member profiles as candidates.

Abstract: Techniques for joint weight attribution for weights of candidate features of a candidate search are described in an example embodiment, disclosed is a system that obtains one or more suggested candidate documents based on a search query specifying one or more parameters. Additionally, the system extracts query intents from the one or more suggested candidate documents, the one or more query intents corresponding to the one or more parameters. Moreover, the system ranks the one or more suggested candidate documents based on the extracted query intents. Furthermore, the system displays top ranked documents on a display device. The system then receives feedback regarding the displayed top ranked documents. Then, weights of a hidden intent are attributed to a set of possible intents based on the received feedback. The feedback can be received retrospectively and proactively. For example, some embodiments perform joint weight attribution based on retrospective and proactive feedback ingestion.