홈 › Papers

The comparison of Wiktionary thesauri transformed into the machine-readable format

Wiktionary is a unique, peculiar, valuable and original resource for natural language processing (NLP). The paper describes an open-source Wiktionary parser: its architecture and requirements followed by a description of Wiktionary features to be taken into account, some open problems of Wiktionary and the parser. The current implementation of the parser extracts the definitions, semantic relations, and translations from English and Russian Wiktionaries. The paper's goal is to interest researchers (1) in using the constructed machine-readable dictionary for different NLP tasks, (2) in extending the software to parse 170 still unused Wiktionaries. The comparison of a number and types of semantic relations, a number of definitions, and a number of translations in the English Wiktionary and the Russian Wiktionary has been carried out. It was found that the number of semantic relations in the English Wiktionary is larger by 1.57 times than in Russian (157 and 100 thousands). But the Russian Wiktionary has more "rich" entries (with a big number of semantic relations), e.g. the number of entries with three or more semantic relations is larger by 1.63 times than in the English Wiktionary. Upon comparison, it was found out the methodological shortcomings of the Wiktionary.

Code (1)

Similar Papers 제목 키워드 기반