Method and system for searching text

• US 6,598,047 B1
• Filed: 07/26/2000
• Issued: 07/22/2003
• Est. Priority Date: 07/26/1999
• Status: Expired due to Fees

• Current Assignee: David W. Russell, John Oborne
• Original Assignee: David W. Russell, John Oborne
• Inventors: Russell, David W., Oborne, John
• Primary Examiner(s): Vu, Kim
• Assistant Examiner(s): Liang, Gwen
• Application Number:
  US09/625,520
• Time in Patent Office: 1,091 Days
• Field of Search: 707/5, 707/103.R, 707/2, 707/3, 704/10, 704/9, 706/16, 706/15, 705/50
• US Class Current: 1/1
• CPC Class Codes: G06F 16/30 of unstructured textual data document management systems G06F16/93
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access
• CPC Class Codes: Y10S 707/99944 Object-oriented database structure

Abstract

An associative search methodology is presented whereby words, phrases, entire documents or technical jargon are input to the system for the purposes of searching a database for like items. A self-organizing associative data structure utilizing Latent Semantic Analysis (LSA) combined with a Hierarchical Mixture of Experts (HME) neural network learning system tracks relationships between words and word groupings as well as user significance feedback to automatically expand the search so that data elements can be found which cross technological/taxonomic boundaries. This results in an increase in both recall and precision parameters for each individual user'"'"'s search.

Claims

1. A method of searching text, the method comprising the following steps, but not necessarily in the order shown:

(A) storing a plurality of documents in a database;

(B) performing a latent semantic analysis of the documents;

(C) generating a latent semantic analysis matrix for the documents;

(D) performing a latent semantic analysis for a search parameter;

(E) calculating a conceptual point for the search parameter;

(F) selecting documents within a predetermined conceptual distance from the conceptual point; and

(G) prioritizing the selected documents using a hierarchical mixture of experts neural network system;

wherein the hierarchical mixture of experts neural network system includes at least one gating network adapted to determine whether data from at least one expert network is used in prioritizing the documents.

2. The method of claim 1 wherein the step of performing a latent semantic analysis for the documents comprises forming a matrix, wherein one axis of the matrix comprises terms in the documents, the other axis of the matrix comprises the documents and the values in the matrix are the frequency of occurrence of the terms.

3. The method of claim 1 wherein feedback from a user is used to prioritize document.

4. The method of claim 1 wherein the hierarchical mixture of experts is initially untrained.

5. The method of claim 1 wherein the hierarchical mixture of experts is trained by user feedback.

6. The method of claim 1 wherein the hierarchical mixture of experts is trained by user feedback, wherein the user feedback is provided by a user making a decision to select a document.

7. The method of claim 6 wherein a user can view selected documents at a plurality of levels of informational detail.

8. The method of claim 7 wherein selecting a document at higher levels of informational detail gives increasing amounts of feedback to the hierarchical mixture of experts.

9. The method of claim 1 wherein the hierarchical mixture of experts is preprogrammed.

10. The method of claim 1 wherein the hierarchical mixture of experts comprises a hierarchical mixture of experts of a first user;

and wherein the hierarchical mixture of experts of the first user comprises the hierarchical mixture of experts of a second user.

11. The method of claim 1 wherein the hierarchical mixture of experts comprises a combination of the hierarchical mixture of experts of multiple users.

12. The method of claim 1 wherein the hierarchical mixture of experts comprises at least one from a group consisting of conceptual nodes and expert nodes.

13. The method of claim 12 further comprising adding a new expert node in response to a new concept in a search.

14. The method of claim 12 further comprising adding a new conceptual node in response to a new concept in a search.

15. The method of claim 12 further comprising recursively expanding an expert node into a set of sub-experts nodes.

16. The method of claim 12 further comprising removing unneeded terms from at least one from a group consisting of conceptual nodes and expert nodes.

17. A method of searching text, the method comprising, but not necessarily in the order shown:

(A) storing a plurality of documents in a database;

(B) analyzing the documents with a neural network;

(C) analyzing a search parameter using a neural network;

(D) calculating a conceptual point for the search parameter;

(E) selecting documents within a predetermined conceptual distance from the conceptual point; and

(F) prioritizing the selected documents using a hierarchical mixture of experts neural network system;

wherein the hierarchical mixture of experts neural network system includes at least one gating network adapted to determine whether data from at least one expert network is used in prioritizing the documents.

18. The method of claim 17 wherein feedback from a user is used to prioritize documents.

19. The method of claim 17 wherein feedback from a user is used to train the hierarchical mixture of experts.

20. The method of claim 17 wherein the hierarchical mixture of experts is initially untrained.

21. The method of claim 17 wherein the hierarchical mixture of experts is preprogrammed.

22. The method of claim 17 wherein the hierarchical mixture of experts comprises at least one from a group consisting of conceptual nodes and expert nodes.