Acquiring dominant compound terms to build Korean domain knowledge bases

Compound terms should be well ranked to reduce laborious work for building domain knowledge bases such as term dictionary and thesaurus. Especially, dominant terms in recent years are valuable in the aspects of coverage and reference. We adopt linguistic filtering using a part-of-speech filter and four combination rules to extract Korean compound terms. Domain seed terms are used to obtain their related terms from the above extracted term list. Term ranking, which considers the dominance trend of terms from several year data, assigns term dominance values to the related terms. Experimental results show that our ranking scheme adequately distributes extracted terms than term frequency ordering to reduce the effort of building domain knowledge bases in the manner of term clustering in three groups; growing, declining, and steady.