A Way to Assign Parts-of-Speech Information to Chinese Frequent Strings

Abstract

A CFS is a frequently used combination of Chinese characters which have been defined in our previous research [11]. A CFS may be a proper noun, like “網際網路” (the Internet), a verb phrase, like “全力動員投入” (try one’s best to mobilize), and so on. If a CFS can have some kinds of POS (part-of-speech), we can use it in more applications. In this paper we propose a method to assign the part-of–speech information to CFSs. If a CFS s is also a word w, we can assign the POSs of w to s. When s is a combination of several words, we will try to find some possible POSs associated with it. We use the Sinica Treebank which contains 38,725 parsing trees as our training and testing corpus. We extract 15,946 parsing rules from 90% of the 38,725 parsing trees. There is 10% of the corpus left for outside test. The accuracies of outside test of assigning POSs to CFSs are 71.02% and 98.81% for top 1 and top 5 choices, respectively.