Evaluating peculiar lexicon for medical record sections identification

We propose a methodology for semi-automatic derivation of knowledge from medical records by means of both statistical and lexical approaches. Moreover, we propose a statistic based methodology for the peculiar lexicon extracted quality evaluation. The evaluation is performed by means of a semantic distance, based on chi-square statistical measure, between the lexicon extracted and the corpus composed by the set of medical record analyzed. The methodology can be used for automatic medical record sections identification in order to improve the interactions among different actors belonging to the health care domain.