Abstract: A method and apparatus of a device that performs a multi-domain query search is described. In an exemplary embodiment, the device receives a query prefix from a client of a user. The device further determines a plurality of search completions across the plurality of separate search domains. In addition, the device ranks the plurality of search completions based on a score calculated for each of the plurality of search completions determined by a corresponding search domain, where at least one of the plurality of search completions is used to generate a plurality of search results without an indication from the user and in response to receiving the query prefix.

Abstract: Methods and systems are disclosed using linear expressions for machine learning (ML) models for ranking search results. In one example, a method for a computer trains a ML model into a decision tree for ranking search results. The decision tree is converted into a linear expression including Boolean terms. The linear expression is transmitted to one or more search computers that use the linear expression to rank search results for a search query. According to another example, a method for a computer having a search engine includes receiving a linear expression including Boolean terms representing a decision tree. The search engine processes a search query and uses the linear expression to rank search results for the search query.

Abstract: Methods and systems that create domain influence scores that can be used to rank or sort search results are described. In one embodiment, a domain influence scoring system begins by preselecting a subset of the domains and allocating an initial influence score to only the domains in the subset while all other domains have an initial influence score of zero. Then links to pages are counted to update the influence scores of each domain. Further, one or more blacklists can be used to modify updating of the influence scores.

Abstract: A device implementing a system for determining whether a semantically relevant answer can be provided with respect to a new question includes a processor configured to identify a question and a semantically relevant answer from source data, and to identify a semantically irrelevant answer to the question from a corpus of data corresponding to the source data based at least in part on the question and the answer. The processor is configured to assign a positive label to the semantically relevant answer, and a negative label to the semantically irrelevant answer. The processor is configured to generate a machine learning model based on the question, the positive label assigned to the semantically relevant answer, and the negative label assigned to the semantically irrelevant answer, and to provide the machine learning model to facilitate a determination of whether a semantically relevant answer can be provided with respect to a subsequent question.

Abstract: The subject technology provides a search index that maps clusters of semantically similar phrases to documents that contain any one of the phrases of the respective cluster. The subject technology may identify the phrases from a set of documents, such as a document corpus, where each of the documents is associated with a document identifier. The subject technology may generate the clusters of semantically similar phrases from the identified phrases, where each of the generated clusters is assigned a cluster identifier. The subject technology generates an index that stores each respective cluster identifier of each respective cluster in association with each document identifier of each of the documents that includes at least one of the phrases contained in the respective cluster. Further, the subject technology stores the index in a memory such that the index may be subsequently utilized to identify documents that match a search query.

Abstract: A device implementing a system for determining whether a semantically relevant answer can be provided with respect to a new question includes a processor configured to identify a question and a semantically relevant answer from source data, and to identify a semantically irrelevant answer to the question from a corpus of data corresponding to the source data based at least in part on the question and the answer. The processor is configured to assign a positive label to the semantically relevant answer, and a negative label to the semantically irrelevant answer. The processor is configured to generate a machine learning model based on the question, the positive label assigned to the semantically relevant answer, and the negative label assigned to the semantically irrelevant answer, and to provide the machine learning model to facilitate a determination of whether a semantically relevant answer can be provided with respect to a subsequent question.

Abstract: The subject technology provides a search index that maps clusters of semantically similar phrases to documents that contain any one of the phrases of the respective cluster. The subject technology may identify the phrases from a set of documents, such as a document corpus, where each of the documents is associated with a document identifier. The subject technology may generate the clusters of semantically similar phrases from the identified phrases, where each of the generated clusters is assigned a cluster identifier. The subject technology generates an index that stores each respective cluster identifier of each respective cluster in association with each document identifier of each of the documents that includes at least one of the phrases contained in the respective cluster. Further, the subject technology stores the index in a memory such that the index may be subsequently utilized to identify documents that match a search query.

Abstract: Methods and systems are disclosed using linear expressions for machine learning (ML) models for ranking search results. In one example, a method for a computer trains a ML model into a decision tree for ranking search results. The decision tree is converted into a linear expression including Boolean terms. The linear expression is transmitted to one or more search computers that use the linear expression to rank search results for a search query. According to another example, a method for a computer having a search engine includes receiving a linear expression including Boolean terms representing a decision tree. The search engine processes a search query and uses the linear expression to rank search results for the search query.

Abstract: Methods and systems that create domain influence scores that can be used to rank or sort search results are described. In one embodiment, a domain influence scoring system begins by preselecting a subset of the domains and allocating an initial influence score to only the domains in the subset while all other domains have an initial influence score of zero. Then links to pages are counted to update the influence scores of each domain. Further, one or more blacklists can be used to modify updating of the influence scores.

Abstract: Author-term affinity scores enable search engines to score and rank search results based on a combination of author (domain) influence and anchor text scoring. An anchor text is scored based on the influence of the domains with which it is associated, including the source domain in which the anchor text is cited and the destination domain to which the anchor text is linked. A contribution of each term to an anchor text's score is determined based on how frequently the term occurs overall. Author-term affinity scores are derived from the anchor text scores and saved for use in subsequent searches to rank search results. During searches an author-query affinity based on the previously stored author-term affinity scores enable a search engine to rank search results to improve relevance such that pages of authors having the most influence are presented before pages of authors having less influence.

Abstract: A method and apparatus of a device that performs a multi-domain query search is described. In an exemplary embodiment, the device receives a query prefix from a client of a user. The device further determines a plurality of search completions across the plurality of separate search domains. In addition, the device ranks the plurality of search completions based on a score calculated for each of the plurality of search completions determined by a corresponding search domain, where at least one of the plurality of search completions is used to generate a plurality of search results without an indication from the user and in response to receiving the query prefix.