Semantic Models for Style-Based Text Clustering

Abstract

The paper addresses some roles of concept-based representations in document clustering to support knowledge discovery. Computational Intelligence algorithms can benefit from semantic networks in the definition of similarity between pairs of documents. After analyzing the tuning of semantic networks in a systematic fashion, the research defines and evaluates a novel semantic-based metrics, which integrates both classical and style-related features of texts. Experimental results confirm the effectiveness of the approach, showing that applying a refined semantic representation into a clustering engine yields consistent structures for information retrieval and knowledge acquisition.