Learning Logic Models for Automated Text Categorization

This work addresses a logical approach to text categorization inside a framework aimed at full automatic paper document processing. The logic representation of sentences required by the adopted learning algorithm is obtained by detecting structure in raw text trough a parser. A preliminary experimentation proved that the logic approach is able to capture the semantics underlying some kind of sentences, even if the assessment of the efficiency of such a method, as well as a comparison with other related approaches, has still to be carried out.