Russian-Language Thesauri: Automatic Construction and Application for Natural Language Processing Tasks

详细

The paper overviews the existing digital Russian-language thesauri and the methods of their automatic construction and application. The authors have analyzed the main characteristics of thesauri published in open access for scientific research, evaluated trends of their development, and their effectiveness in solving natural language processing tasks. Statistical and linguistic methods of thesaurus construction that allow automation of their development and reduce the labor costs of expert linguists have been studied. In particular, algorithms for extracting keywords and semantic thesaurus relations of all types have been considered and the quality of the thesauri generated with the use of these tools was assessed. To illustrate features of various methods of constructing thesaurus relations, the authors developed a combined method that fully automatically generates a specialized thesaurus based on a text corpus of a selected domain and several existing linguistic resources. The proposed method was used to conduct experiments on two Russian-language text corpora that represent two different domains: articles on migration and tweets. The resulting thesauri were analyzed by means of an integrated assessment that had been developed by the authors in a previous study and allows one to determine various aspects of the analyzed thesaurus and appraise the quality of the methods of its generation. The analysis revealed the main advantages and disadvantages of various approaches to thesaurus construction and extraction of semantic relations of different types, and also made it possible to identify potential focus areas for future research.