Query suggestions based on search data

摘要

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating query suggestions based on search data. In one aspect, a method includes receiving, by one or more computers, a first query, determining query refinements based on the first query, generating, from the query refinements, refinement clusters, each refinement cluster corresponding to a particular topic and each refinement cluster including query refinements that are determined to belong to the particular topic to which the refinement cluster corresponds, ranking the refinement clusters, and selecting the refinement cluster that is highest in the ranking relative to other refinement clusters in the ranking as a first search refinement cluster for the first query.

主权项

1. A method comprising:
determining, by one or more computers, query refinements for a first query; generating, from the query refinements, refinement clusters, each refinement cluster corresponding to a particular topic included in the queries in the refinement cluster and not included in the first query and each refinement cluster including query refinements that are determined to belong to the particular topic to which the refinement cluster corresponds; scoring each of the refinement clusters using a quantity of unique n-grams associated with the particular topic of the refinement cluster that occur in the queries included in the refinement cluster, wherein, for each of the refinement clusters, a score for the refinement cluster counts each of the unique n-grams once; ranking the refinement clusters based on the scores associated with the refinement clusters, the refinement cluster that is highest in the ranking relative to the other refinement clusters having a highest score of the refinement clusters; selecting the refinement cluster that is highest in the ranking relative to other refinement clusters in the ranking as a first search refinement cluster for the first query; generating first topic prompt data based on the topic associated with the first search refinement cluster, the first topic prompt data describing a question that comprises a request for a first user input of an n-gram that is a subtopic of the topic associated with the first search refinement cluster; receiving, for a user session, the first query; providing, for the user session, the first topic prompt data in response to receipt of the first query; receiving, for the user session, data indicating a particular n-gram that is i) a subtopic of the topic associated with the first search refinement cluster and ii) an answer to the topic prompt data, the receipt of the data indicating the particular n-gram subsequent to the receipt of the first query without other intervening user input for the user session; and providing, for the user session, search results responsive to the first query and the particular n-gram.