A Definition-based Shared-concept Extraction within Groups of Chinese Synonyms: A Study Utilizing the Extended Chinese Synonym Forest

Synonym groups can serve as resourceful linguistic metadata for information extraction and word sense disambiguation. Nevertheless, the reasons two words can be categorized into a particular synonym group need further study, especially when no explanation is available as to why any two words are synonymous. Lexical resources, such as the Chinese Synonym Forest (or Tongyici Cilin) (Mei et al. 1983), assemble lexical items into hierarchical categories via manual categorization. Other than this, statistical measures, such as co-existing probability, have been adopted widely to verify synonymous relationships. Nevertheless, a purely statistical method does not provide description that can help interpret why such a synonymous relationship occurs. We propose a novel method for the study of shared concepts within any synonym group by comparing co-existing words in the dictionary definition of each member in the group. The co-existing words are seen as the representatives of shared concepts that can be used for interpretating any hidden meaning among members of a synonym group. We also compare our results with the thesaurus function in the Sketch Engine (Kilgarriff et al. 2004), which uses statistical data in the form of Sketch scores. The results show that our method can produce concept words according to dictionary definitions, but this method also has its limitations, as it works only with a finite number of synonyms and under limited computing resources.