Connotative Meaning of Military Chat Communications

摘要

Automatic processing of military chat text in operational environments will be necessary to provide automated data collection, collation, and usage for tactical updates, post-mission operational analysis, and watch turnover. The informal nature of chat communications allows the relay of far more information than the technical content of messages. This AFRL in-house project combined components of the methodology applied in a Syracuse University project for IARPA's AQUAINT program with additional research activities to analyze databases of military chat. The project proposed to conduct a study of how humans recognize connotative cues expressing uncertainty, perception of personal threat, and urgency; formulate linguistic and non-linguistic means for recognizing such cues; develop algorithms to automatically perform that recognition, and evaluate the prototype recognition algorithms. The project built a matrix of speech 'cues' representative of uncertainty, perception of personal threat, and urgency, but also applied maximum entropy analysis and a combined rule-based/statistical algorithm. Recall, precision and F-score measures for each methodology were determined.