GAITHERSBURG, MD, 03/02/2004 -- Lockheed Martin [NYSE: LMT] announced today a new version of its AeroText^TM software, a cutting-edge information extraction tool that solves problems with data overload. Version 3.0 improves data extraction, processes a variety of document formats and has been enhanced to support more accurate time-line analysis.

AeroText^TM version 3.0 enhancements improve the effectiveness of data extraction in a number of key areas: processing of new document formats is now added to the current ASCII and Unicode support including Adobe® PDF, Microsoft® Word^TM and Excel^TM formats; temporal expressions have been normalized to coordinated universal time to support more accurate time-line analysis; the geospatial-disambiguation function is enhanced to improve recognition of ambiguous place names in a document; and relationship extraction is enhanced to provide finer resolution of entity relationships.

The AeroText development environment now has integrated accuracy metrics for evaluating the extraction of complex relationships. Using AeroText^TM version 3.0, a knowledge worker can compare rule additions and modifications over time to evaluate the combined effectiveness of entity and relationship-level extraction and measure slot-level extraction.

Lockheed Martin developed the AeroText^TM software as a foundation solution for problems with data overload and data access in many fields, such as biomedical research, market research, competitive assessments and national security. The software finds relevant information, such as entities, relationships and events, automatically in natural language text documents of all major formats. AeroText^TM significantly reduces the time required for information exploration by quickly identifying information with an accuracy that matches or exceeds human interpretation. It also discovers subtle relationships among data elements that would otherwise be unavailable to a human investigator.

The AeroText^TM Core Rules Knowledge Base identifies entity information such as people, organizations, locations, time and phone numbers as well as relationships among these entities. Users can employ - without change - the AeroText^TM Core Rules Knowledge Base for standard extraction tasks. For demands of a specific subject area or for more distinctive data, such as a foreign language, AeroText^TM supports the smooth upgrade or development of unique and tailored rules knowledge bases.

The flexibility offered by AeroText^TM in the rule-development arena leads to constant improvements in extraction accuracy and to a customized solution regardless of the human language or application subject matter.