Generalized Representation of Syntactic Structures

Analysis of language provides important insights into the underlying psychological properties of individuals and groups. While the majority of language analysis work in psychology has focused on semantics, psychological information is encoded not just in what people say, but how they say it. In the current work, we propose Conversation Level Syntax Similarity Metric-Group Representations (CASSIM-GR). This tool builds generalized representations of syntactic structures of documents, thus allowing researchers to distinguish between people and groups based on syntactic differences. CASSIM-GR builds off of Conversation Level Syntax Similarity Metric by applying spectral clustering to syntactic similarity matrices and calculating the center of each cluster of documents. This resulting cluster centroid then represents the syntactical structure of the group of documents. To examine the effectiveness of CASSIM-GR, we conduct three experiments across three unique corpora. In each experiment, we calculate the clustering accuracy and compare our proposed technique to a bag-of-words approach. Our results provide evidence for the effectiveness of CASSIM-GR and demonstrate that combining syntactic similarity and tf-idf semantic information improves the total accuracy of group classification.