Sequential Data Mining for Information Extraction from Texts

This paper shows the benefit of using data mining methods for Biological Natural Language Processing. A method for discovering linguistic patterns based on a recursive sequential pattern mining is proposed. It does not require a sentence parsing nor other resource except a training data set. It produces understandable results and we show its interest in the extraction of relations between named entities. For the named entities recognition problem, we propose a method based on a new kind of patterns taking account the sequence and its context. MOTS-CLÉS : extraction d’information, fouille de données, motifs séquentiels et motifs LSR, TAL appliqué aux textes biologiques et génétiques.