Mapping an automated survey coding task into a probabilistic text categorization framework

This paper describes how to applya probabilistic Text Categorization method to a different and new domain where documents are answers to open end questionnaires and codes viewed as categories consist of a hierarchical model. A reduced size training set mayb e used taking advantage of the hierarchical organization of categories. The system developed in this framework aims at helping psychologists in the evaluation of open end surveys inquiring about job candidates' competencies.