Ambiguity Resolution for Vt-N Structures in Chinese

Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 928–937, October 25-29, 2014, Doha, Qatar. c⃝2014 Association for Computational Linguistics Ambiguity Resolution for Vt-N Structures in Chinese Yu-Ming Hsieh1,2 Jason S. Chang2 Keh-Jiann Chen1 1 Institute of Information Science, Academia Sinica, Taiwan 2 Department of Computer Science, National Tsing-Hua University, Taiwan morris@iis.sinica.edu.tw, jason.jschang@gmail.com kchen@iis.sinica.edu.tw Abstract The syntactic ambiguity of a transitive verb (Vt) followed by a noun (N) has long been a problem in Chinese parsing. In this paper, we propose a classifier to resolve the ambiguity of Vt-N structures. The design of the classifier is based on three important guidelines, namely, adopting linguistically motivated features, using all available resources, and easy in- tegration into a parsing model. The lin- guistically motivated features include semantic relations, context, and morpho- logical structures; and the available re- sources are treebank, thesaurus, affix da- tabase, and large corpora. We also pro- pose two learning approaches that resolve the problem of data sparseness by auto- parsing and extracting relative knowledge from large-scale unlabeled data. Our experiment results show that the Vt-N classifier outperforms the cur- rent PCFG parser. Furthermore, it can be easily and effectively integrated into the PCFG parser and general statistical pars- ing models. Evaluation of the learning approaches indicates that world knowledge facilitates Vt-N disambigua- tion through data selection and error cor- rection. 1 Introduction In Chinese, the structure of a transitive verb (Vt) followed by a noun (N) may be a verb phrase (VP), a noun phrase (NP), or there may not be a dependent relation, as shown in (1) below. In general, parsers may prefer VP reading because a transitive verb followed by a noun object is nor- mally a VP structure. However, Chinese verbs can also modify nouns without morphological inflection, e.g., 養殖/farming 池/pond. Conse- quently, parsing Vt-N structures is difficult be- cause it is hard to resolve such ambiguities with- out prior knowledge. The following are some typical examples of various Vt-N structures: 1) 解決/solve 問題/problem  VP 解決/solving 方案/method  NP 解決/solve 人類/mankind (問題/problem)None To find the most effective disambiguation fea- tures, we need more information about the Vt-N  NP construction and the semantic relations between Vt and N. Statistical data from the Sini- ca Treebank (Chen et al., 2003) indicates that 58% of Vt-N structures are verb phrases, 16% are noun phrases, and 26% do not have any de- pendent relations. It is obvious that the semantic relations between a Vt-N structure and its con- text information are very important for differen- tiating between dependent relations. Although the verb-argument relation of VP structures is well understood, it is not clear what kind of se- mantic relations result in NP structures. In the next sub-section, we consider three questions: What sets of nouns accept verbs as their modifi- ers? Is it possible to identify the semantic types of such pairs of verbs and nouns? What are their semantic relations? 1.1 Problem Analysis Analysis of the instances of NP(Vt-N) structures in …

Content truncated to 500 words