Semantic flow in language networks discriminates texts by genre and publication date

摘要

We propose a framework to characterize documents based on their semantic flow. The proposed framework encompasses a network-based model that connected sentences based on their semantic similarity. Semantic fields are detected using standard community detection methods. As the story unfolds, transitions between semantic fields are represented in Markov networks, which in turn are characterized via network motifs (subgraphs). Here we show that different book characteristics (such as genre and publication date) are discriminated by the adopted semantic flow representation. Remarkably, even without a systematic optimization of parameters, philosophy and investigative books were discriminated with an accuracy rate of 92.5%. While the objective of this study is not to create a text classification method, we believe that semantic flow features could be used in traditional network-based models of texts that capture only syntactical/stylistic information to improve the characterization of texts. (C) 2020 Elsevier B.V. All rights reserved.