Text Categorization for Generation of a Historical Shipbuilding Ontology.

This paper deals with the task of developing a text corpus for the automatic generation of a historical shipbuilding domain ontology. Standard methods of analysis produce unsatisfactory results due to the limited nomenclature of available texts and lexical evolution of language. In this work, a parser developed by authors is used for lemmatization and word-sense disambiguation. The parser is based on an external classifier and provides the unambiguous relationship between each lexeme and class. The documents are represented as vectors in the topic space. The experiments show that the proposed method of categorization produces results very close to the expert opinion and at the same time is sufficiently resistant to the historical dynamics of the vocabulary.