Chinese-Korean Word Alignment Using Contrastive Analysis and Concept Similarity

摘要

In a different language pair in terms of linguistic typology and genealogy, such as Chinese and Korean, a conventional method of word alignment does not effectively work and produce low-quality results because of dissimilarity of alignment units. This paper redefines the word alignment unit through contrastive analysis of the two languages to improve recall and precision. And also a two-phase scoring mechanism of concept similarity is proposed to improve the coverage of monolingual thesaurus. Only linguistic knowledge such as similarity of lexical structure and concept similarity is employed in this paper, so a small-sized and general bilingual corpus can also be used to complete word alignment without the traditional statistical information. For the experiment, we randomly extracted 200 Chinese-Korean sentence pairs from the Joins Newspaper, and achieved 82.5% precision and 62.8% recall.