×

Determining user intent from query patterns

  • US 8,868,548 B2
  • Filed: 07/22/2010
  • Issued: 10/21/2014
  • Est. Priority Date: 07/22/2010
  • Status: Active Grant
First Claim
Patent Images

1. A computer-implemented method, comprising:

  • identifying a query submission in a session, the query including one or more query terms, and obtaining a plurality of search results responsive to the query, each search result identifying a respective document that is associated with a respective score indicating a relevance of the document to the query;

    associating a plurality of intents with the session based on the one or more query terms including an intent refining query term, wherein the intent refining query term was previously submitted with a plurality of different queries a quantity of times that exceeds a threshold;

    identifying one or more of the respective documents, wherein each of the identified documents was selected by users in a plurality of sessions associated with at least one of the intents with a higher or lower frequency than when the identified document was selected by users in a plurality of sessions not associated with the at least one intent;

    for each of the identified documents;

    identifying, for each of the plurality of intents, an intent boost that represents a score adjustment for the identified document based on the intent;

    determining, for each of the plurality of intents, a confidence value that is based on a quantity of the plurality of sessions associated with the intent; and

    modifying the respective score of the document based on a document boost factor, the document boost factor being based on a function of the intent boost for each of the plurality of intents and the confidence value for each of the plurality of intents; and

    ranking the plurality of search results based on the respective scores, including the one or more modified respective scores.

View all claims
  • 2 Assignments
Timeline View
Assignment View
    ×
    ×