Domain Lexico-Taxonomy: an Approach Towards Multi-domain Language

This paper deals with the domain barrier issues in language processing. Our work centers on Domain Lexico-Taxonomy (DLT), a domain taxonomy enhanced by domain lexicons. We propose DLT as an infrastructure for crossing domain barriers. By using DLT with WordNet and Domain Taxonomy, we can get 15160 Chinese lemmas in 463 domains. We estimate the accuracy of five domain’s lemmas, and get 89.74 % in average. Look from all lemmas, each lemma is assigned 1.38 domains in average. By gathering the web pages which Google query returned as testing data, we make some experiments to confirm effectiveness of domain lexicons. Finally, we analyze the result to prove the usefulness of domain lexicons obtained by the DLT approach. Key word: multi-domain language processing, domain lexicon, domain taxonomy, WordNet.