System and method of finding documents related to other documents and of finding related words in response to a query to refine a search

• US 9,064,005 B2
• Filed: 07/12/2007
• Issued: 06/23/2015
• Est. Priority Date: 05/09/2001
• Status: Expired due to Fees

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Stensmo, Jan Magnus
• Primary Examiner(s): Reyes, Mariela
• Assistant Examiner(s): Black, Linh
• Application Number:
  US11/776,634
• Publication Number:
  US 20080016050A1
• Time in Patent Office: 2,903 Days
• Field of Search: 707 5- 6, 707/713, 707765-766
• US Class Current: 1/1
• CPC Class Codes: G06F 16/3325 Reformulation based on results of preceding query
• CPC Class Codes: G06F 16/3346 using probabilistic model
• CPC Class Codes: Y10S 707/99932 Access augmentation or optimizing
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching
• CPC Class Codes: Y10S 707/99934 Query formulation, input preparation, or translation
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access

Abstract

A computer-implemented system and method is disclosed for retrieving documents using context-dependant probabilistic modeling of words and documents. The present invention uses multiple overlapping vectors to represent each document. Each vector is centered on each of the words in the document and includes the local environment. The vectors are used to build probability models that are used for predictions of related documents and related keywords. The results of the statistical analysis are used for retrieving an indexed document, for extracting features from a document, or for finding a word within a document. The statistical evaluation is also used to evaluate the probability of relation between the key words appearing in the document and building a vocabulary of key words that are generally found together. The results of the analysis are stored in a repository. Searches of the data repository produce a list of related documents and a list of related terms. The user may select from the list of documents and/or from the list of related terms to refine the search and retrieve those documents which meet the search goal of the user with a minimum of extraneous data.

Claims

1. A method, comprising:

receiving at least one search term from a user;

automatically determining additional search terms related to the at least one search term at least in part by performing a probability calculation determining a relevance of the additional search terms to the at least one search term using a Simple Bayes probability model developed from analysis of at least one known document, wherein the analysis involves generating a plurality of overlapping vectors corresponding to a plurality of windows centered on a plurality of respective words of the at least one known document by counting a number of occurrences of the plurality of respective words within each window, wherein each respective word on which a window is centered is excluded from a vector generated corresponding to the window, and wherein the probability calculation is performed using at least one processor; and

providing at least one of the additional search terms to the user.

2. The method of claim 1, further comprising providing the user a list of documents automatically determined to be relevant to the at least one search term, the list comprising the at least one known document.

3. The method of claim 2, further comprising receiving a selection of the at least one additional search term from the user, searching documents based on the at least one search term and the at least one additional search term, generating a list of documents automatically determined to be relevant to the at least one search term and the at least one additional search term using the simple Bayes probability model executing on the at least one processor, and providing to the user a subset of the list of documents automatically determined to be relevant to the at least one search term and the at least one additional search term, the subset excluding the at least one known document.

4. The method of claim 1, wherein using the Simple Bayes probability model comprises calculating Mutual Information between the at least one search term and at least one of the additional search terms and calculating bias coefficients representative of how common the at least one search term is in the at least one known document.

5. The method of claim 4, wherein calculating Mutual Information and bias coefficients comprises modeling conditional probability as a sum of pair-wise mutual information values.

6. The method of claim 2, further comprising retrieving at least one document of the list of documents.

13. The method of claim 1, wherein the plurality of windows comprises a respective window centered on each word of the at least one known document.

7. A computer usable storage medium having computer-executable instructions stored thereon, the computer-executable instructions configured to cause a computer to implement a method when executed, the method comprising:

for at least one search term received from a user, automatically determining additional search terms related to the at least one search term at least in part by performing a probability calculation determining a relevance of the additional search terms to the at least one search term using a Simple Bayes probability model developed from analysis of at least one known document, wherein the analysis involves generating a plurality of overlapping vectors corresponding to a plurality of windows centered on a plurality of respective words of the at least one known document by counting a number of occurrences of the plurality of respective words within each window, wherein each respective word on which a window is centered is excluded from a vector generated corresponding to the window; and

providing at least one of the additional search terms to the user.

8. The computer usable storage medium of claim 7, wherein using the Simple Bayes probability model comprises calculating Mutual Information between the at least one search term and at least one of the additional search terms and calculating bias coefficients representative of how common the at least one search term is in the at least one known document.

9. The computer usable storage medium of claim 8, wherein the Mutual Information and bias coefficients are calculated by modeling conditional probability as a sum of pair-wise mutual information values.

11. The computer usable storage medium of claim 7, wherein the method further comprises providing to the user a list of documents automatically determined to be relevant to the at least one search term, the list comprising the at least one known document.

12. The computer usable storage medium of claim 11, wherein the method further comprises, in response to receiving a selection of the at least one of the additional search terms from the user, searching documents based on the at least one search term and the at least one additional search term, generating a list of documents automatically determined to be relevant to the at least one search term and the at least one additional search term using the Simple Bayes probability model, and providing the user a subset of the list of documents automatically determined to be relevant to the at least one search term and the at least one additional search term, the subset excluding the at least one known document.

14. The computer usable storage medium of claim 7, wherein the plurality of windows comprises a respective window centered on each word of the at least one known document.

10. A processor implemented system, comprising:

a first module executing on a processor and configured to, for at least one search term received from a user, automatically determine additional search terms related to the at least one search term at least in part by performing a probability calculation determining a relevance of the additional search terms to the at least one search term using a Simple Bayes probability model developed from analysis of at least one known document, wherein the analysis involves generating a plurality of overlapping vectors corresponding to a plurality of windows centered on a plurality of respective words of the at least one known document by counting a number of occurrences of the plurality of respective words within each window, wherein each respective word on which a window is centered is excluded from a vector generated corresponding to the window; and

a second module configured to provide at least one of the additional search terms to the user.

15. The processor implemented system of claim 10, wherein the plurality of windows comprises a respective window centered on each word of the at least one known document.