POINT PAPER ON DOCUMENT CONTENT ANALYSIS & RETRIEVAL SYSTEM (DCARS)

DCARS Summary:

- A concept-based text retrieval system
- Developing for intelligence analysts at the National Air Intelligence Center (NAIC), WPAFB
- Uses ConQuest, a high quality Commercial-off-the-Shelf (COTS) product, as text retrieval engine

DCARS Goals:

- To reduce the time & effort it takes analysts to find relevant documents & data from text
- To reduce the volume of irrelevant documents retrieved (high precision)
- To be adaptable to new domains (topic areas)
- To incorporate "lower risk" technology, to maximize chances of helping analysts

How DCARS Supports the Warfighter

- Bomb Damage Assessments (BDAs) & Quick Response Tasks (QRTs)

-- Reduces time it takes to find the data required for analyses; means more time for analysis
-- Improves the quality of analyses provided to decision-makers, since more time for analysis
-- DCARS could be deployed operationally to find data for BDAs & QRTs real-time, in field

- A variety of search methods, including:

-- Natural-language (concept-based) retrieval : 250,000 concepts already
-- Statistical
-- Boolean/Keyword

-- Query by example (using sample text)
-- Document summaries (to be developed)
-- Relevancy ranking of documents

-- Facilitate query formation & refinement
-- Facilitate interpretation & comparison of query results

-- ConQuest's Core General English Dictionary is domain independent
-- Developing tool to automatically build dictionaries for new/specialized domains
-- Developing dictionary extendibility tool to add new words to dictionary

-- ConQuest's system scored highly at '92, '93, and '95 Text Retrieval Conference (TREC)
-- ConQuest is a COTS product already in use in several locations

Contractor: Calspan/Buffalo NY 14225/Program Mgr./(716) 631-6844 Lab Program Mgr: AFRL/IFE/(315) 330-4518, DSN 587-4518/29 Apr 96

  Last Reviewed:    22 October 2002
pagemaster email