Improving Classification of Multi-Lingual Web Documents using Domain Ontologies

In this paper, we deal with the problem of analyzing and classifying web documents to several major categories/classes in a given domain using domain ontology. We present the ontology-based web content mining methodology that contains such main stages as collecting a training set of labeled documents from a given domain, building a classification model above this domain given the domain ontology, and classification of new documents via the induced model. We tested the proposed methodology in a specific domain, namely web pages containing information about production of certain chemicals. Using our methodology, we are interested to identify all relevant web documents while ignoring the documents that do not contain any relevant information. Our system receives as input an OWL file built in Protege tool, which contains the domain-specific ontology, and a set of web documents classified by a human expert as ”relevant” or ”non-relevant”. We use a language-independent key-phrase extractor with integrated ontology parser (defined in a given language) for creating the database from input documents and use it as a training set for the classification algorithm. The system classification accuracy using various levels of ontology is evaluated.The current version of our system supports web content mining in English, Arabic, Russian, and Hebrew languages.