Ontology Based Machine Learning for Semantic Multiclass Classification

Following the development of semantic web technologies, many ontologies and thesauri have been proposed to index resources during the last decade. However, despite their expressiveness, those knowledge models do not always cover all the points of interest within dedicated applications. Therefore, alternative ad hoc taxonomies have been developed to support resources classifying processes. This paper proposes a method that bridges existing knowledge models with ad hoc taxonomies to address the problem of textual documents classification. Usually, documents are indexed according to different knowledge models: keywords, thesauri, ontologies. Nevertheless, for a project leader, additional information are needed to organize documents. In response to a particular need of one of our partners, we have developed a learning method based on the use of ontologies for modelling a semantic classification process. This method allows the expert user to match their needs by optimising text document classification.