Textual Analysis and Linking of Narratives (TALON)

Abstract

This paper describes the design and implementation of a methodology for parsing free-text narratives in an expert decision system for law enforcement. The problem arises from local law enforcement agencies' inability to effectively utilize criminal incident data for crime prevention. The paper proposes previously unsupported functionality in criminal incident data association expert decision systems - the automated ability of incorporating information from narrative fields from a criminal incident record management system into the analysis. A software module called TALON, the Textual Analysis and Linking of Narratives implements the methodology. Performance evaluation utilizes precision and recall metrics from information retrieval in a controlled experiment. This experiment uses the methodology on a set of documents wherein the relevance of every document is known for every possible selection of a target search document. The work presented provides a strong basis for measuring the utility of implementing alternative forms of narrative parsing for similar analyses.