Browse Prior Art Database

A Method of Enhancing Results of Full Text Search by Applying Structured Query Filters.

IP.com Disclosure Number: IPCOM000016069D
Original Publication Date: 2002-Jul-06
Included in the Prior Art Database: 2003-Jun-21
Document File: 4 page(s) / 84K

Publishing Venue

IBM

Abstract

Post Disclosure Text Drawings

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 38% of the total text.

Page 1 of 4

  A Method of Enhancing Results of Full Text Search by Applying Structured Query Filters.

Post Disclosure Text & Drawings

Enter any additional information relating to this disclosure below:

Title: A Method of Enhancing Results of Full Text Search by Applying Structured Query Filters.

Traditional full text search is based on the index of words, extracted from a set of unstructured documents, such as HTML pages, etc. The search results for given query terms depend on different characteristics of the documents, included in the index, e.g. the number of query terms in a given document, the proximity of query terms, and others.

Query

Search Client

Engine

index

 Query Processor

 Result Processor

Result

One of the frequently encountered problems with text search is the need to filter search results, based on attributes not related directly to the query terms, like creation/modification date, author name, etc. Search engines help to resolve this problem by supporting limited sets of predefined meta tags that contain some frequently used attributes of documents. IBM's DB2 Net Search Extender provides complementary functionality to a traditional search engine by offering numeric search on a range of values. For example, searching for books sold within a specific price range or published within a specific year range. Existing solutions have obvious limitations, because they can support only limited number of predefined filters based on meta tags or numeric fields.

This invention presents a new method of enhancing results of full text search by exploring the structured information of documents as a complementary information for the search engine's search results. This allows to filter full text search results by combining them with results of arbitrary query on the structured information of the same set of documents. This method is based on the component library and the searchable components in the library that are described the disclosure "Method of Dynamic Document Generation on Web Environment" (YOR8-2001-1224).

Problem Statement

To improve text search result, the traditional full text search engine needs to filter the text search results based on some criteria. In existing search engines the criteria are based on some predefined attributes or meta-tags like creation/modification date, author name, etc. These criteria are limited and restricted to predefined or frequently used attributes of documents.

How Is Problem Solved

This invention is a new way of using the structured information of documents to generalize the criteria for filtering the search results of the full text search engines.

1

[This page contains 1 picture or other non-text object]

Page 2 of 4

The following figure shows the overview diagram of the invention. The client component sends a natural language query to the Advanced Query Processor which uses a NL processing tool to analyze the received query, splits the original query into two queries - full text search query and struc...