Deep Exploration and Filtering of Text (DEFT)
Abstract
The Deep Exploration and Filtering of Text (DEFT) program will enable automated extraction, processing, and inference of information from text in operationally relevant application domains. A key DEFT emphasis is to determine the implied and hidden meaning in text through probabilistic inference, anomaly detection, and disfluency analysis. To accomplish this, DEFT will develop and apply formal representations for basic facts, spatial, temporal, and associative relationships, causal and process knowledge, textually entailed information, and derived relationships and correlated actions/events. DEFT inputs may be in English or in a foreign language and sources may be completely free-text or semi-structured reports, messages, documents, or databases. DEFT will extract knowledge at scale for open source intelligence and threat analysis. Planned transition partners include the intelligence community and operational commands.
Document Details
- Document Type
- Accomplishment
- Publication Date
- Oct 01, 2015
- Source ID
- 12245ad5385cb9c31533bfdc51067877