Structuring and Fusing Text
Abstract
Much information important for battlefield assessments is transmitted as unstructured text, including both unclassified documents available to the general public as well as highly classified reports and messages. Text can convey such highly valued information as adversary plans and goals. Unfortunately, valuable text nuggets may be buried in massive amounts of less important information, and may be difficult to find. Once found, different aspects of an entity or activity may be scattered among different text sources, making it difficult to assemble into a coherent picture able to convey context, relationships, and trends. This paper describes a methodology for structuring and fusing open source information so that it may be presented on maps and diagrams. This process employs formal ontologies for the fusion domain and for evidential reasoning, a commercial tool for text extraction and structuring, and tools to help operators review, edit, and augment the fusion products. It features a fusion pedigree to document the audit trail of sources and processes contributing to a fusion product. The fusion steps are: 1) collect structured and unstructured information related to the entity or event of interest; 2) extract and structure free text; 3) create event reports from each structured record, whether derived from free text or previously structured sources; 4) create ontology-based communication reports from these records, 5) associate these reports, 6) cue manual search for additional information, and 7) fuse the information.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 2006
- Accession Number
- ADA474153
Entities
People
- David F. Noble