Abstract: A method for annotating images to create a corpus for training a multi-task computer vision machine learning model is presented. The method comprises receiving, at one or more annotation specialist models, a plurality of images to be annotated. Via operation of the one or more annotation specialist models, pre-filtered annotations are generated for the plurality of images. Via operation of a data filtering and enhancement module, the pre-filtered annotations are filtered in accordance with predefined noise criteria so as to output candidate annotations for the plurality of images. The method further comprises, for each of one or more candidate annotations, selectively (1) storing the candidate annotation into the corpus as a final annotation for its associated image, or (2) adding the candidate annotation to its associated image using the one or more annotation specialist models and the data filtering and enhancement module for subsequent iterative annotation and filtering.

Abstract: Techniques for artificially generating commentary for online content including news items. In an aspect, a personification engine incorporates a machine learning model trained using corpus elements comprising an item of online content and relevant commentary. The personification engine is configured to generate relevant commentary when provided with an item of online content such as a news item. In a further aspect, a chatbot engine incorporates a model similarly trained using corpus element comprising a comment and a relevant response. The chatbot engine is configured to generate relevant responses to user comments in the context of a forum or comments section related to the item of online content.

Abstract: Technologies relating to generation of training data for use in training a computer-implemented ranker are described. A product name extracted from a URL in a log entry of a search log of a search engine is mapped to a catalog product name from a catalog of a website, wherein the catalog product name identifies a product that is available for acquisition by way of a webpage of a website. A query assigned to the URL in the log entry is assigned to the catalog product name due to the product name being mapped to the catalog product name. The computer-implemented ranker is trained based upon the query being assigned to the catalog product name.

Abstract: A method and apparatus for providing recommended contents. The method comprises: A search query being received; an event being selected from a predetermined event set based on the search query; the recommended contents being provided in a structured template including one or more elements; each element is associated with a feature of the event; each feature of the event corresponds to a search intent. Each element includes one or more sub-contents, and each sub-content is correlated to corresponding recommended contents.

Abstract: A search system that is configured to generate personalized ranked lists of search results for end users is disclosed. The search system provides backend search services to several websites of different organizations, where the websites are configured to offer goods for acquisition to the end users. The search system is configured to construct profiles for users that interact with webpages of the websites, and is further configured to construct personalized re-rankers based upon such profiles.

Abstract: A computer-implemented technique is described herein for facilitating discovery of information items. The technique involves providing at least one set of indicators to a user via a user interface presentation. The indicators represent respective topics. Each distance d between a pair of respective indicators, associated with respective topics (t, t?), is based on at least: a similarity between the topic t and the topic t? as expressed by a knowledge graph; and a dynamic and personalized joint measure of the user's current interest in the topic t and the topic t?. The technique further involves providing at least one information item to the user in response to the user's engagement with the user interface presentation.

Abstract: Techniques for improving content load time are described. A method comprises: receiving, from a search service, a plurality of search results in response to a search query, wherein each of the plurality of search results includes a result link pointing to an object stored on a content source; receiving, from the search service, auxiliary information indicating that, for at least one of the plurality of search results, a copy of at least a portion of the object that the result link of the at least one search result points to is cached within a cache service associated with the search service; and retrieving, from the cache service, the cached copy of at least the portion of the object in response to a selection of the result link of the at least one search result.

Abstract: Technologies relating to generation of training data for use in training a computer-implemented ranker are described. A product name extracted from a URL in a log entry of a search log of a search engine is mapped to a catalog product name from a catalog of a website, wherein the catalog product name identifies a product that is available for acquisition by way of a webpage of a website. A query assigned to the URL in the log entry is assigned to the catalog product name due to the product name being mapped to the catalog product name. The computer-implemented ranker is trained based upon the query being assigned to the catalog product name.

Abstract: A search system that is configured to generate personalized ranked lists of search results for end users is disclosed. The search system provides backend search services to several websites of different organizations, where the websites are configured to offer goods for acquisition to the end users. The search system is configured to construct profiles for users that interact with webpages of the websites, and is further configured to construct personalized re-rankers based upon such profiles.

Abstract: A search system is described herein. The search system is configured to search over several different catalogs, where each catalog includes entries that represent goods and/or services offered for acquisition by companies. The search system receives a query and outputs search results from a catalog based upon the query. The search system includes a general-purpose ranker that is configured to receive the search results and rank the search results, thereby generating a ranked list of search results. The general-purpose ranker is used to rank the search results regardless of the catalog from which the search results were retrieved. The search system further includes a re-ranker that is customized for the catalog, wherein the re-ranker receives the ranked list of search results and re-ranks such search results, thereby generating a second ranked list of search results.

Abstract: A method and apparatus for providing recommended contents. The method comprises: A search query being received; an event being selected from a predetermined event set based on the search query; the recommended contents being provided in a structured template including one or more elements; each element is associated with a feature of the event; each feature of the event corresponds to a search intent. Each element includes one or more sub-contents, and each sub-content is correlated to corresponding recommended contents.

Abstract: Techniques for improving content load time are described. A method comprises: receiving, from a search service, a plurality of search results in response to a search query, wherein each of the plurality of search results includes a result link pointing to an object stored on a content source; receiving, from the search service, auxiliary information indicating that, for at least one of the plurality of search results, a copy of at least a portion of the object that the result link of the at least one search result points to is cached within a cache service associated with the search service; and retrieving, from the cache service, the cached copy of at least the portion of the object in response to a selection of the result link of the at least one search result.

Abstract: Aspects of the present disclosure relate to systems and methods for sharing content. More specifically, a user may select a subset of content from displayed content. The selection may be made using a visual indicator. Once the subset of content is selected, the subset of content is analyzed to determine how much content of the subset of content the user wants to share. Once the determination is complete, a content card containing the subset of content is generated and provided to the user. The user may then add additional content to the content card and electronically share the content card by posting the content card on a social media site.

Abstract: Techniques for artificially generating commentary for online content including news items. In an aspect, a personification engine incorporates a machine learning model trained using corpus elements comprising an item of online content and relevant commentary. The personification engine is configured to generate relevant commentary when provided with an item of online content such as a news item. In a further aspect, a chatbot engine incorporates a model similarly trained using corpus element comprising a comment and a relevant response. The chatbot engine is configured to generate relevant responses to user comments in the context of a forum or comments section related to the item of online content.

Abstract: A computer-implemented technique is described herein for facilitating discovery of information items. The technique involves providing at least one set of indicators to a user via a user interface presentation. The indicators represent respective topics. Each distance d between a pair of respective indicators, associated with respective topics (t, t?), is based on at least: a similarity between the topic t and the topic t? as expressed by a knowledge graph; and a dynamic and personalized joint measure of the user's current interest in the topic t and the topic t?. The technique further involves providing at least one information item to the user in response to the user's engagement with the user interface presentation.

Abstract: A computer-implemented method for processing user entered query data to improve results of a search of pages using a database, when searching the internet, is disclosed. The method includes receiving the user entered query data and parsing each word of the query data and segmenting words using probability to determine a likelihood that the word is for a particular name. And, associating the particular names with a name tag to create one or more tagged name terms. Then, normalizing each of the tagged name terms and the normalizing including boosting information if found in the database and determining proximity between selected ones of the tagged name terms. The method then generates an optimized search query that incorporates normalized terms and operators. The optimized search query being applied to the internet to enable search results to be produced and displayed to the user in response to the entered query data.

Abstract: In one embodiment, access a search query comprising one or more query words, at least one of the query words representing one or more query concepts; access a network document identified for a search query by a search engine, the network document comprising one or more document words, at least one of the document words representing one or more document concepts; semantic-text match the search query and the network document to determine one or more negative semantic-text matches; and construct one or more negative features based on the negative semantic-text matches.

Abstract: A computer-implemented method for processing user entered query data to improve results of a search of pages using a database, when searching the internet, is disclosed. The method includes receiving the user entered query data and parsing each word of the query data and segmenting words using probability to determine a likelihood that the word is for a particular name. And, associating the particular names with a name tag to create one or more tagged name terms. Then, normalizing each of the tagged name terms and the normalizing including boosting information if found in the database and determining proximity between selected ones of the tagged name terms. The method then generates an optimized search query that incorporates normalized terms and operators. The optimized search query being applied to the internet to enable search results to be produced and displayed to the user in response to the entered query data.

Abstract: Techniques for determining when and how to transform words in a query to its plural or non-plural form in order to provide the most relevant search results while minimizing computational overhead are provided. A dictionary is generated based upon the words used in a specified number of previous most frequent search queries and comprises lists of transformations from plural to singular and singular to plural. Unnecessary transformations are removed from the dictionary based upon language modeling. The word to transform is determined by finding the last non-stop re-writable word of the query. The context of the transformed word is confirmed in the search documents and a version of the query is executed using both the original form of the word and the transformation of the word.

Abstract: One particular embodiment accesses a first set of search queries comprising one or more first search queries; extracts one or more features based on the first set of search queries, trains a search-query classifier using the features; accesses a second search query provided by a user; determines whether the second search query has implicit and general local intent using the search-query classifier; if the second search query has implicit and general local intent, then determines a location associated with the user; and identifies a search result in response to the second search query based at least in part on the location associated with the user; and presents the search result to the user.