Contrasting Machine Learning Approaches for Microtext Classification

摘要

The goal is classification of microtext: classifying lines of military chat, or posts, which contain items of interest. This paper evaluates non-linear statistical data modeling techniques, and compares with our previous results using several text categorization and feature selection methodologies. The chat posts are examples of 'microtext', or text that is generally very short in length, semi-structured, and characterized by unstructured or informal grammar and language. These three distinct attributes cause different results than traditional long-form free text. In this paper, we further characterize microtext. Highly accurate classification of microtext entries is crucial to facilitate more complex information extraction. Although this study focused specifically on tactical updates via chat, we believe the findings are applicable to content of a similar linguistic structure regardless of domain. This includes other microtext sources such as IM/XMPP, SMS, voice transcriptions, and micro-blogging such as Twitter(tm).