UTH: SVM-based Semantic Relation Classification using Physical Sizes

Proceedings of the 4th International Workshop on Semantic Evaluations (SemEval-2007), pages 464–467, Prague, June 2007. c⃝2007 Association for Computational Linguistics UTH: Semantic Relation Classification using Physical Sizes Eiji ARAMAKI Takeshi IMAI Kengo MIYO The University of Tokyo Hospital departmentKazuhiko OHE 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan aramaki@hcc.h.u-tokyo.ac.jp Abstract Although researchers have shown increas- ing interest in extracting/classifying seman- tic relations, most previous studies have ba- sically relied on lexical patterns between terms. This paper proposes a novel way to accomplish the task: a system that captures a physical size of an entity. Experimental results revealed that our proposed method is feasible and prevents the problems inherent in other methods. 1 Introduction Classification of semantic relations is important to NLP as it would benefit many NLP applications, such as machine translation and information re- trieval. Researchers have already proposed various schemes. For example, Hearst (1992) manually de- signed lexico-syntactic patterns for extracting is-a relations. Berland and Charniak (1999) proposed a similar method for part-whole relations. Brin (1998) employed a bootstrapping algorithm for more spe- cific relations (author-book relations). Kim and Baldwin (2006) and Moldovan et al.(2004) focused on nominal relations in compound nouns. Turney (2005) measured relation similarity between two words. While these methods differ, they all utilize lexical patterns between two entities. Within this context, our goal was to utilize infor- mation specific to an entity. Although entities con- tain many types of information, we focused on the physical size of an entity. Here, physical size refers to the typical width/height of an entity. For example, we consider book to have a physical size of 20×25 cm, and book to have a size of 10×10 m, etc. We chose to use physical size for the following reasons: 1. Most entities (except abstract entities) have a physical size. 2. Several semantic relations are sensitive to phys- ical size. For example, a content-container rela- tion (e1 content-container e2) naturally means that e1 has a smaller size than e2. A book is also smaller than its container, library. A part- whole relation has a similar constraint. Our next problem was how to determine physi- cal sizes. First, we used Google to conduct Web searches using queries such as “book (*cm x*cm)” and “library (*m x*m)”. Next, we extracted numeric expressions from the search results and used the av- erage value as the physical size. Experimental results revealed that our proposed approach is feasible and prevents the problems in- herent in other methods. 2 Corpus We used a corpus provided by SemEval2007 Task #4 training set. This corpus consisted of 980 anno- tated sentences (140 sentences×7 relations). Table 1 presents an example. Although the corpus contained a large quantity of information such as WordNet sense keys, comments, etc., we used only the most pertinent information: entity1 (e1), entity2 (e2), and its relation (true/false) 464 The <e1>library</e1> contained <e2>books </e2> of guidance on the processes. WordNet(e1) = "library\%1:14:00::", WordNet(e2) = "book\%1:10:00::", Content-Container(e2, e1) = "true", Query = "the * contained books" Table 1: An Example …

Content truncated to 500 words