Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for related search suggestions for products. In one aspect, a method includes receiving a query for searching documents wherein the query specifies a first product. A number of occurrences of the first product in a plurality of prior queries of the documents submitted by users is determined. A product category for the first product is determined based on one or more attributes associated with the first product, the category having a predetermined plurality of category attributes. A suggested product is selected wherein the suggested product is associated with attributes that satisfy the category attributes.

Abstract: Methods, systems and apparatus, including computer programs encoded on a computer storage medium, for disambiguating names in a document corpus. In an aspect, a method includes generating context term lists for a person name, each context term list being a list of context terms from a resource for the person name; clustering the context term lists into a plurality of clusters, each of the clusters of context term lists including context term lists that are most similar to the cluster relative to other clusters; for each of the clusters, selecting a representative term for the cluster; receiving the person name as a search query; and generating a plurality of query suggestions from the search query and the representative terms for the clusters, each query suggesting being a combination of the person name and one representative term.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for related search suggestions for products. In one aspect, a method includes receiving a query for searching documents wherein the query specifies a first product. A number of occurrences of the first product in a plurality of prior queries of the documents submitted by users is determined. A product category for the first product is determined based on one or more attributes associated with the first product, the category having a predetermined plurality of category attributes. A suggested product is selected wherein the suggested product is associated with attributes that satisfy the category attributes.

Abstract: Methods, systems, and apparatus, including computer program products, for generating query refinements using query components. In general, one aspect features a method that includes the acts of receiving a query comprising a plurality of terms; identifying first and second components of the query, wherein each component comprises one or more of the terms of the query and the components do not share a term from the query, and wherein the first component appears before the second component in the query; determining, for each component, a plurality of different respective component refinements; and combining the component refinements to create a plurality of query refinements for the query, including combining a first component refinement for the first component with a second component refinement for the second component to create a query refinement so that the first component refinement appears before the second component refinement in the query refinement.

Abstract: Methods, systems, and apparatus, including computer program products, for generating query refinements from user preference data. A group of query pairs are obtained. Each query pair includes a first query and a second query. A quality score is determined for each query pair from user preference data for documents responsive to both the first and the second query. A diversity score is determined for each query pair having a quality score satisfying a quality threshold, the diversity score determined from user preference data for documents responsive to the second, but not the first, query. For each query pair having a quality score satisfying the quality threshold and a diversity score satisfying a diversity threshold, the second query of the query pair is associated with the first query of the query pair as a candidate refinement for the first query.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for determining word boundary likelihoods in potentially incomplete text. In one aspect, a method includes selecting query sequences from the query, each query sequence being at least a portion of a word n-gram, the word n-gram being a subsequence of up to n words selected from the second sequence of words of the query, and for each query sequence: determining one or more query sequence keys for the query sequence; determining at least one of a word boundary count and a non-word boundary count for each query sequence key, each word-boundary count and non-word boundary count being dependent on the context of the query sequence; and associating, in a data storage device, the at least one word boundary count and non-word boundary counts with each query sequence key.

Abstract: Systems, methods, and apparatus, including computer program products, for collaborative filtering are provided. A method is provided. The method includes clustering a plurality of entities with respect to one or more latent variables in a probability distribution model of a relationship between a set of entities and a set of items, the probability distribution model comprising a probability distribution of the set of items with respect to the latent variables. The method also includes, as new items are added to the set of items, updating the probability distribution of the set of the items with respect to the latent variables, and generating an updated relationship score for an entity with respect to the set of items based on the entity's fractional membership in the clustering with respect to the latent variables and based on the updated probability distribution of the set of the items with respect to the latent variables.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for obtaining query completions. In one aspect, a method includes receiving a query input in a search engine query input field in a user interface. The method also includes submitting the query input as a first query stem to an autocompletion module. The method also includes receiving a first response from the autocompletion module, the first response providing no first query autocompletions. The method also includes submitting a second query stem to the autocompletion module, the second query stem being the first query stem with a first prefix removed. The method also includes receiving a second response from the autocompletion module including one or more second autocompletions satisfying a second quality test. The method also includes providing second autocompletions for presentation on the user interface.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for determining word boundary likelihoods in potentially incomplete text. In one aspect, a method includes selecting query sequences from the query, each query sequence being at least a portion of a word n-gram, the word n-gram being a subsequence of up to n words selected from the second sequence of words of the query, and for each query sequence: determining one or more query sequence keys for the query sequence; determining at least one of a word boundary count and a non-word boundary count for each query sequence key, each word-boundary count and non-word boundary count being dependent on the context of the query sequence; and associating, in a data storage device, the at least one word boundary count and non-word boundary counts with each query sequence key.

Abstract: Methods, systems, and apparatus, including computer program products, for generating query refinements from user preference data. A group of query pairs are obtained. Each query pair includes a first query and a second query. A quality score is determined for each query pair from user preference data for documents responsive to both the first and the second query. A diversity score is determined for each query pair having a quality score satisfying a quality threshold, the diversity score determined from user preference data for documents responsive to the second, but not the first, query. For each query pair having a quality score satisfying the quality threshold and a diversity score satisfying a diversity threshold, the second query of the query pair is associated with the first query of the query pair as a candidate refinement for the first query.

Abstract: Methods, systems and apparatus, including computer programs encoded on a computer storage medium, for disambiguating names in a document corpus. In an aspect, a method includes generating context term lists for a person name, each context term list being a list of context terms from a resource for the person name; clustering the context term lists into a plurality of clusters, each of the clusters of context term lists including context term lists that are most similar to the cluster relative to other clusters; for each of the clusters, selecting a representative term for the cluster; receiving the person name as a search query; and generating a plurality of query suggestions from the search query and the representative terms for the clusters, each query suggesting being a combination of the person name and one representative term.

Abstract: Methods, systems, apparatus, including computer program products, for providing fresh related search suggestions in response to a user submitted query are presented. In one implementation, a plurality of prior queries are selected wherein each of the prior queries was submitted as a search query a number of times during a recent time period and satisfies a criterion. For each of the prior queries, the prior is selected as a candidate query based on one or more of: a determination that search results responsive to the prior query include a number of news results that satisfy a second threshold, and relevance data indicative of user behavior relative to the search results responsive to the prior query. In response to receiving a user query, one or more candidate queries are selected that match the user query.

Abstract: Methods, systems, and apparatus, including computer program products, for generating query refinements. In one implementation, a method includes dividing a search query into one or more n-grams; determining that each of the one or more n-grams represents a same concept, and in response to the determination: selecting one or more candidate refinement queries; determining one or more categories for the search query; determining one or more categories for each of the candidate refinement queries; determining a respective first score for each of the candidate refinement queries; and identifying as query refinements for the search query one or more of the candidate refinement queries based on the first scores.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for identifying query refinements from sibling queries. In one aspect, a method includes associating each of a plurality of parent queries with a respective group of one or more child queries for the parent query, identifying one or more candidate sibling queries for a particular child query, selecting one or more final sibling queries for the particular child query from the one or more candidate sibling queries, and associating the final sibling queries with the particular child query as query refinements.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for related search suggestions for products. In one aspect, a method includes receiving a query for searching documents wherein the query specifies a first product. A number of occurrences of the first product in a plurality of prior queries of the documents submitted by users is determined. A product category for the first product is determined based on one or more attributes associated with the first product, the category having a predetermined plurality of category attributes. A suggested product is selected wherein the suggested product is associated with attributes that satisfy the category attributes.

Abstract: Methods, systems, and apparatus, including computer program products, for generating query refinements from user preference data. A group of query pairs are obtained. Each query pair includes a first query and a second query. A quality score is determined for each query pair from user preference data for documents responsive to both the first and the second query. A diversity score is determined for each query pair having a quality score satisfying a quality threshold, the diversity score determined from user preference data for documents responsive to the second, but not the first, query. For each query pair having a quality score satisfying the quality threshold and a diversity score satisfying a diversity threshold, the second query of the query pair is associated with the first query of the query pair as a candidate refinement for the first query.

Abstract: Methods, systems, apparatus, including computer program products, for providing fresh related search suggestions in response to a user submitted query are presented. In one implementation, a plurality of prior queries are selected wherein each of the prior queries was submitted as a search query a number of times during a recent time period and satisfies a criterion. For each of the prior queries, the prior is selected as a candidate query based on one or more of: a determination that search results responsive to the prior query include a number of news results that satisfy a second threshold, and relevance data indicative of user behavior relative to the search results responsive to the prior query. In response to receiving a user query, one or more candidate queries are selected that match the user query.

Abstract: Systems, methods, and apparatus, including computer program products, for collaborative filtering are provided. In one implementation, a computer-implemented method is provided. The method includes receiving a shard of data representing a subset of a set of entities and a subset of a set of items, generating an iteration of a maximum likelihood estimate of a probability distribution model of a relationship between the set of entities and the set of items, the probability distribution model comprising a probability distribution of the set of items with respect to latent variables and a probability distribution of the latent variables with respect to the set of users, and generating statistics from results from the generating step which are passed to different shards for use in a next iteration of the maximum likelihood estimate.

Abstract: Systems, methods, and apparatus, including computer program products, for collaborative filtering are provided. A method is provided. The method includes clustering a plurality of entities with respect to one or more latent variables in a probability distribution model of a relationship between a set of entities and a set of items, the probability distribution model comprising a probability distribution of the set of items with respect to the latent variables. The method also includes, as new items are added to the set of items, updating the probability distribution of the set of the items with respect to the latent variables, and generating an updated relationship score for an entity with respect to the set of items based on the entity's fractional membership in the clustering with respect to the latent variables and based on the updated probability distribution of the set of the items with respect to the latent variables.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for identifying query refinements from sibling queries. In one aspect, a method includes associating each of a plurality of parent queries with a respective group of one or more child queries for the parent query, identifying one or more candidate sibling queries for a particular child query, selecting one or more final sibling queries for the particular child query from the one or more candidate sibling queries, and associating the final sibling queries with the particular child query as query refinements.