Method And System For Adaptively Weighting Full Text Search Results Of Structured Data (17-Dec-2009)

Method And System For Adaptively Weighting Full Text Search Results Of Structured Data

Many full text search algorithms exist for searching unstructured data-like web pages. They utilize clever techniques beyond just word frequencies to deliver accurate results. When applied to structured data, however, search engines do not adaptively weight search hits based on the importance of a field in a structured document or database. Some allow you to explicitly set the importance of a field to weight search results, but this is a manual process. This also requires knowledge of the document or database schema and must be updated manually as the schema changes.

This technique can be applied to any structured data, regardless of the actual document or database structure, and field importance assessments can be continuously re-evaluated and updated as the document or database structure changes.

This analysis of structured data can be broken down into three broad categories:

1. Field Definition: How a field is defined in a database schema, including field length and field type

2. Field Presentation: How fields appear in a form as part of an application's user interface
3. Historical Use: Analysis of existing data to determine how a field is used

Field Definition

Simply looking at field metadata reveals clues about how important a particular field might be. For instance, search hits for fields that have shorter maximum character lengths should be weighted higher than multi-line fields with longer character lengths. In a change management system, a defect's "summary" or "product" field values (fewer characters) might be weighted higher than "description" or "comments" (more characters).

Fields of certain types might be weighted higher. For instance, a field that references another record could be weighted higher than a field that represents a date. System fields such as "DBI...