Search By Document Type And Relevance

• US 20070150473A1
• Filed: 05/16/2006
• Published: 06/28/2007
• Est. Priority Date: 12/22/2005
• Status: Abandoned Application

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Xu, Jun, Li, Hang, Cao, Yunbo
• Application Number:
  US11/383,638
• Publication Number:
  US 20070150473A1
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques

Abstract

A method of finding documents. A method of finding documents comprising, ranking documents according to relevance to form a ranked relevance list, ranking documents according to type to form a ranked type list, and combining the ranked relevance list and the ranked type list to form a list of documents ranked by relevance and type.

Claims

1. A method of searching by document type comprising:

ranking documents according to relevance to form a ranked relevance list;

ranking documents according to type to form a ranked type list; and

combining the ranked relevance list and the ranked type list to form a list of documents ranked by relevance and type.

2. The method of searching by document type of claim 1 further comprising learning a relevance model from a set of relevance model training data.

3. The method of searching by document type of claim 2 in which the relevance model set of training data includes a query, a document, and a label.

4. The method of searching by document type of claim 1 further comprising learning a type model from a set of type training data.

5. The method of searching by document type of claim 4 in which the type model set of training data includes a document and a label.

6. The method of searching by document type of claim 1 in which interpolation is performed by linear combination.

7. The method of searching by document type of claim 1 in which interpolation is performed by thresholding.

8. The method of searching by document type of claim 1 in which ranking documents according to relevance to form a ranked relevance list is performed by a document relevance search.

9. The method of searching by document type of claim 8 in which the document relevance search is Okapi.

10. The method searching by document type of claim 1 in which ranking documents according to type to form a ranked type list is performed by a classifier.

11. The method searching by document type of claim 10 in which the classifier is logistic regression.

12. A computer readable media encoded to perform a typed search comprising:

performing a typed search to produce a first result and a second result; and

combining the first result and the second result.

13. The computer readable media encoded to perform a typed search of claim 12 in which combining is performed by linear combination.

14. The computer readable media encoded to perform a typed search of claim 12 in which combining is performed by thresholding.

15. The computer readable media encoded to perform a typed search of claim 12 in which the typed search includes utilizing a type model.

16. The computer readable media encoded to perform a typed search of claim 12 in which the typed search includes utilizing a relevance model.

17. A system for searching by document type comprising:

a means for determining a relevance model producing a first result;

a means for determining a type model for producing a second result; and

a means for combining the first result and the second result.

18. The system for searching by document type of claim 17 in which the means for combining includes a means for linearly combining the first result and the second result.

19. The system for searching by document type of claim 17 in which the means for combining includes a means for thresholding the first result and the second result.

20. The system for searching by document type of claim 17 in which an instruction document is found.