Finding Taxonomical Relation from an MRD for Thesaurus Extension

R. Dale et al. (Eds.): IJCNLP 2005, LNAI 3651, pp. 357 – 365, 2005. © Springer-Verlag Berlin Heidelberg 2005 Finding Taxonomical Relation from an MRD for Thesaurus Extension SeonHwa Choi and HyukRo Park Dept. of Computer Science, Chonnam National University, 300 Youngbong-Dong, Puk-Ku Gwangju, 500-757, Korea csh123@dreamwiz.com, hyukro@chonnam.ac.kr Abstract. Building a thesaurus is very costly and time-consuming task. To alleviate this problem, this paper proposes a new method for extending a thesaurus by adding taxonomic information automatically extracted from an MRD. The proposed method adopts a machine learning algorithm in acquiring rules for identifying a taxonomic relationship to minimize human-intervention. The accuracy of our method in identifying hypernyms of a noun is 89.7%, and it shows that the proposed method can be successfully applied to the problem of extending a thesaurus. 1 Introduction As the natural language processing (NLP) systems became large and applied to wide variety of application domains, the need for a broad-coverage lexical knowledge-base has increased more than ever before. A thesaurus, as one of these lexical knowledge- bases, mainly represents a taxonomic relationship between nouns. However, because building broad-coverage thesauri is a very costly and time-consuming job, they are not readily available and often too general to be applied to a specific domain. The work presented here is an attempt to alleviate this problem by devising a new method for extending a thesaurus automatically using taxonomic information extracted from a machine readable dictionary (MRD). Most of the previous approaches for extracting hypernyms of a noun from the definition in an MRD rely on the lexico-syntactic patterns compiled by human experts. Not only these methods require high cost for compiling lexico-syntactic patterns but also it is very difficult for human experts to compile a set of lexical-syntactic patterns with a broad-coverage because, in natural languages, there are various different expressions which represent the same concept. Accordingly the applicable scope of a set of lexico-syntactic patterns compiled by human is very limited. To overcome the drawbacks of human-compiled lexico-syntactic patterns, we use part-of-speech (POS) patterns only and try to induce these patterns automatically using a small bootstrapping thesaurus and machine learning methods. The rest of the paper is organized as follows. We introduce the related works in section 2. Section 3 deals with the problem of features selection. In section 4, our problem is formally defined as a machine learning method and discuss implementation details. Section 5 is devoted to experimenal result. Finally, we come to the conclusion of this paper in section 6. 358 S.H. Choi and H.R. Park 2 Related work [3] introduced a method for the automatic acquisition of the hyponymy lexical relation from unrestricted text, and gave several examples of lexico-syntactic patterns for hyponymy that can be used to detect these relationships including those used here, along with an algorithm for identifying new patterns. Hearst’s approach is complementary to statistically based approaches that find semantic relations between terms, in that hers requires a single specially expressed instance of a relation while the others require …

Content truncated to 500 words