Exécution de notre outil de synthèse sur un nouvel article

Vérifiez l’article ci-dessous, puis lancez sa synthèse.


Text Genre Classification Using only Parts of Speech

Schulman Alan (Middle Tennessee State University), Salvador E. Barbosa (Middle Tennessee State University)

International Conference on Computational Science and Computational Intelligence (CSCI)
Licence : non indiquée (tous droits réservés)

Résumé : This paper discusses a unique approach to English text classification. A topic that the authors have not found to be explicitly explored is the role that parts of speech (POS) have on text classification. This paper aims to answer whether POS patterns exist within a text that can distinguish it from one of a different genre. The experiments were carried out in two phases: using the Natural Language Toolkit, input texts are processed into feature vectors reflecting their POS patterns. These feature vectors are then passed to a separate program that virtually maps them into a coordinate space, ultimately clustering them based on spatial proximity through the non-parametric K-means algorithm. A set of test sentences is then evaluated with the goal of having the cluster centroid to which they are closest possess the same classification. This work showed that text genre, a category based solely on lexical content, correlates to syntactic patterns of the language used.

Publié le 01 Déc. 2018