Automated content processing ofspoken and written discourse: Text coherence, essays, and team analyses : Identifying information and tenor in texts

Latent Semantic Analysis (LSA) is a computational model of human knowledge acquisition and a practical application for information retrieval and concept-based text matching. This paper illustrates ways in which LSA can be applied to performing automated content analyses of text and discourse. Three example domains are discussed: Analyses of the coherence of written documents, analyses of student essays for judging knowledge, and analyses of team discourse for predicting team performance and automatic tagging of discourse content. The results show that LSA can be used in these domains for modeling cognitive and social properties of discourse, as well as provide useful applications that make predictions based on automated analyses of spoken and written discourse.