System and method for blending the results of a classifier and a search engine

• US 20050149504A1
• Filed: 01/07/2004
• Published: 07/07/2005
• Est. Priority Date: 01/07/2004
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Ratnaparkhi, Adwait
• Granted Patent: US 7,424,469 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/3346 using probabilistic model
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching
• CPC Class Codes: Y10S 707/99937 Sorting

Abstract

A method to produce a list of documents retrieved from a search engine, ranked according to their relevance to search terms, includes interleaving the lists resulting from two searches. The first search is one that occurs over documents where the list is ranked using a first probabilistic ranking algorithm. The second list is a list of documents obtained by searching over a query log of prior searches using the same search terms. The second list is ranked according to a second probabilistic algorithm. Interleaving of the two lists includes eliminating duplicate occurrences of documents and appending the remaining documents to the end of the interleaved list if the two lists of individual search results differ in length. The interleaved list may then be presented to a user for review.

Claims

1. A method for a generating an ordered list of documents, the method comprising:

receiving at least one search term;

searching in at least one document database for documents related to the at least one search term;

retrieving a first list of documents relating to the at least one search term and ranking the first list using a first probabilistic ranking algorithm forming a first ranked list;

searching in at least one query log storage for documents related to the at least one search term;

accessing a second list of documents from the query log storage and ranking the second list according to a second probabilistic ranking algorithm forming a second ranked list; and

interleaving the first ranked list and the second ranked list forming an interleaved list of documents.

2. The method of claim 1, further comprising:

providing the interleaved list of documents to a user for review.

3. The method of claim 1, further comprising:

removing any duplicates in the interleaved list.

4. The method of claim 1, wherein searching in at least one document database comprises using a search engine connected to at least one document database via a network.

5. The method of claim 1, wherein searching in at least one query log storage comprises searching in one of a local query log storage or a networked query log storage.

6. The method of claim 1, wherein interleaving to form an interleaved list comprises alternating the occurrence of documents taken from the first ranked list and the second ranked list.

7. The method of claim 1, wherein interleaving comprises generating the interleaved list such that a first listed document of the second ranked list is a first entry in the interleaved list of documents.

8. The method of claim 1, wherein interleaving comprise generating the interleaved list such that a first listed document of the first ranked list is a first entry in the interleaved list of documents.

9. The method of claim 1, wherein if one of the first ranked list and the second ranked list is longer in length than the other, then interleaving further comprises appending a remainder of the longer list to the end of the interleaved list.

10. The method of claim 1, wherein ranking the first list using a first probabilistic ranking algorithm forming a first ranked list comprises using a form of the OKAPI ranking algorithm.

11. The method of claim 1, wherein ranking the second list according to a second probabilistic ranking algorithm forming a second ranked list comprises using one of a Naive Bayes and a Maximum Entropy algorithm.

12. A system for a generating a list of documents, the system comprising:

a processor having access to memory, the memory having instructions which, when executed, perform the method comprising;

receiving at least one search term;

searching in at least one document database for documents related to the at least one search term;

retrieving a first list of documents relating to the at least one search term and ranking the first list using a first probabilistic ranking algorithm forming a first ranked list;

searching in at least one query log storage for documents related to the at least one search term;

accessing a second list of documents from the query log storage and ranking the second list according to a second probabilistic ranking algorithm forming a second ranked list; and

interleaving the first ranked list and the second ranked list forming an interleaved list of documents.

13. The system of claim 12, further comprising:

a network interface for communications with the at least one document database located on the network.

14. The system of claim 12, further comprising:

a network interface for communication with a search engine located on the network.

15. The system of claim 12, further comprising:

a display device wherein the interleaved list may be reviewed.

16. The system of claim 12, wherein the method step of ranking the first list using a first probabilistic ranking algorithm forming a first ranked list comprises using a form of the OKAPI ranking algorithm.

17. The system of claim 12, wherein the method step of ranking the second list according to a second probabilistic ranking algorithm forming a second ranked list comprises using one of a Naive Bayes and a Maximum Entropy algorithm.

18. A computer-readable medium having computer-executable instructions for performing a method for generating a list of documents, the method comprising:

receiving at least one search term;

searching in at least one document database for documents related to the at least one search term;

retrieving a first list of documents relating to the at least one search term and ranking the first list using a first probabilistic ranking algorithm forming a first ranked list;

searching in at least one query log storage for documents related to the at least one search term;

accessing a second list of documents from the query log storage and ranking the second list according to a second probabilistic ranking algorithm forming a second ranked list; and

interleaving the first ranked list and the second ranked list forming an interleaved list of documents.

19. The computer-readable medium of claim 18, wherein the method further comprises:

providing the interleaved list of documents to a user for review.

20. The computer-readable medium of claim 18, wherein the method step of searching in at least one document database for documents related to the at least one search term comprises using a search engine connected to a document database via a network.

21. The computer-readable medium of claim 18, wherein the method step of interleaving comprises alternating the occurrence of documents taken from the first ranked list and the second ranked list.

22. The computer-readable medium of claim 18, wherein the method step of interleaving comprises generating the interleaved list such that a first ranked document of the second ranked list is a first entry in the interleaved list of documents.

23. The computer-readable medium of claim 18, wherein the method step of interleaving comprises generating the interleaved list such that a first listed document of the first ranked list is a first entry in the interleaved list of documents.

24. The computer-readable medium of claim 18, wherein the method step of ranking the first list using a first probabilistic ranking algorithm forming a first ranked list comprises using a form of the OKAPI ranking algorithm.

25. The computer-readable medium of claim 18, wherein the method step of ranking the second list according to a second probabilistic ranking algorithm forming a second ranked list comprises using one of a Naive Bayes and a Maximum Entropy algorithm.