Automatic extracting and extending lexical attributes from chinese machine readable dictionary

This paper proposes a method for automatic extracting and extending lexical attributes from a Chinese Machine Readable Dictionary (MRD). The method acquires lexical attributes such as density, reserve and ductility for the class metal from definitions in MRD, not using any instantiated objects. Compared with previous works, our method takes advantage of hyponymy semantic relation to extend classes conceptually and acquire more attributes. Although the method for acquisition of hyponyms is very simple and only a few hyponyms are acquired, experimental results show that they can help to improve the performance of attribute extraction from MRD effectively.