Enriched search features based in part on discovering people-centric search intent

• US 8,510,322 B2
• Filed: 06/17/2011
• Issued: 08/13/2013
• Est. Priority Date: 06/17/2011
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Ray, Aninda, Meyerzon, Dmitriy, Udupa, Uppinakuduru Raghavendra
• Primary Examiner(s): Wong, Leslie
• Application Number:
  US13/162,620
• Publication Number:
  US 20120323877A1
• Time in Patent Office: 788 Days
• Field of Search: 707/999.003, 707/999.005, 707/999.107, 707/759, 707/765, 707/698
• US Class Current: 707/765
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation

Abstract

A search environment of an embodiment includes name mining and matching features used in part to identify people-centric queries and provide an enriched search experience, but is not so limited. A method of an embodiment operates to provide an expanded query based in part on a geometric similarity measure, an edit distance measure, a string similarity measure, and a cumulative similarity measure. A search system of an embodiment includes a mined candidate generator component and a name matcher component used in part to identify name queries and provide an expanded query that includes original query terms and one or more valid mined names. Other embodiments are also disclosed.

Claims

1. A method comprising:

using a server computer as part of;

generating a number of mined candidates based in part on using a learned hash function and valid pass-through of an original query input that includes determining if the original query input satisfies a first restrictive filter stage comprising an invalid query detector filter and a second restrictive filter stage comprising a minimum query tokens detector filter;

using an active mining filter having a number of phases to further process mined candidates;

using an adaptive filter comprising a multi-level name matcher constraint set that includes first threshold levels and second threshold levels; and

generating an expanded query that includes the original query input and any valid name corresponding to any mined candidate that satisfies all of the first threshold levels or all of the second threshold levels associated with the multi-level name matcher constraint set.

2. The method of claim 1, further comprising generating the expanded query using the multi-level name matcher constraint set including a geometric measure, an edit distance measure, a string similarity measure, and a cumulative similarity measure.

3. The method of claim 2, further comprising generating the expanded query using the multi-level name matcher constraint set, wherein the geometric measure comprises a canonical geometric score, the edit distance measure comprises a Levenshtein edit distance score, the string similarity measure comprises a Jaro-Winkler distance score, and the cumulative similarity measure comprises a cumulative distance score based in part on the canonical geometric score, the Levenshtein edit distance score, and the Jaro-Winkler distance score.

4. The method of claim 1, further comprising automatically executing the expanded query as part of discovering a searching intent to be people-centric including using customized actions to enrich a search experience.

5. The method of claim 4, further comprising automatically executing the expanded query containing a candidate expansion having a highest confidence measure according to a specified threshold to provide one or more person records including using the customized actions to perform one or more of rendering search results in a different way, returning a number of authored documents or other items, and fetching profile information from social and other networks.

6. The method of claim 1, further comprising using a first threshold value and a second threshold value as part of first and second geometric measure assessments, using a first threshold value and a second threshold value as part of first and second edit distance measure assessments, using a first threshold value and a second threshold value as part of first and second string similarity measure assessments, and using a first threshold value and a second threshold value as part of first and second cumulative similarity measure assessments, including returning one or more valid personal names to be used as part of the expanded query if one or more mined candidates satisfy all of the first threshold level values or all of the second threshold level values.

7. The method of claim 6, further comprising using the second threshold values once one of the first threshold level values is not satisfied.

8. The method of claim 5, further comprising assigning different threshold level values based in part on importance of one or more of geometric similarity constraints and structural similarity constraints.

9. The method of claim 1, further comprising using the similarity score and one or more mined name candidates that are nearest neighbors to original query tokens, wherein a similarity score is calculated for each nearest neighbor in part to determine relatedness between each mined first and last name candidate and the original query tokens.

10. The method of claim 9, further comprising mining a list of nearest neighbors for each query term of the original query input and determining if each mined nearest neighbor satisfies a similarity threshold value.

11. The method of claim 1, further comprising exiting to an organic search service if the number of mined candidates is zero or if no mined candidate satisfies all threshold levels of one of the first or second constrain sets of the adaptive filter.

12. A search system including at least one processor, the system comprising:

an invalid query detector component configured to filter invalid queries based in part on a number of original query tokens;

a minimum query tokens detector component configured to filter out query inputs having a defined number of query tokens that are less than a defined query token threshold;

a mined candidate generator component configured to generate mined name candidates based in part on the number of original query tokens and a similarity measure;

a name matcher component configured to generate a number of valid names based in part on an output from the mined candidate generator and a plurality of threshold values associated with a first threshold determination stage and a second threshold determination stage, wherein the name matcher component is configured to generate the number of valid names based in part on an output from the mined candidate generator and a geometric measure, an edit distance measure, a string similarity measure, and a cumulative similarity measure; and

a query expander component configured to provide expanded queries based in part on one or more original query terms and one or more valid names.

13. The search system of claim 12, the mined candidate generator to generate mined name candidates from a name repository based in part on a number of original query terms and a similarity measure threshold value including a selection of all first and last names from a name directory component that include one or more mined terms.

14. The search system of claim 13, the mined candidate generator to eliminate mined personal names that have fewer terms than the number of original query terms.

15. The search system of claim 12, the name matcher component to generate the number of valid names based in part on an output from the mined candidate generator and a canonical geometric score, a Levenshtein edit distance score, a Jaro-Winkler distance score, and a cumulative similarity score.

16. Computer storage, including instructions which, when executed, operate to:

use an original query to mine personal names including using an invalid query detector filter, a minimum query tokens detector filter, and an active mining filter having a number of phases;

generate altered query terms comprising personal names using an adaptive filter including a geometric measure, an edit distance measure, a string similarity measure, and a cumulative similarity measure; and

provide an altered query using one or more valid personal names having satisfied at least one level of the adaptive filter.

17. The computer storage of claim 16, including instructions which, when executed, operate to automatically execute the altered query against a name directory.

18. The computer storage of claim 16, including instructions which, when executed, operate to use the altered query to provide information associated with one or more individuals including authored materials or contact information associated with a social network.