Japanese-Spanish Thesaurus Construction Using English as a Pivot

Japanese-Spanish Thesaurus Construction Using English as a Pivot Jessica Ramírez, Masayuki Asahara, Yuji Matsumoto Graduate School of Information Science Nara Institute of Science and Technology Ikoma, Nara, 630-0192 Japan {jessic-r,masayu-a,matsu}@is.naist.jp Abstract We present the results of research with the goal of automatically creating a multilin- gual thesaurus based on the freely available resources of Wikipedia and WordNet. Our goal is to increase resources for natural language processing tasks such as machine translation targeting the Japanese-Spanish language pair. Given the scarcity of re- sources, we use existing English resources as a pivot for creating a trilingual Japanese- Spanish-English thesaurus. Our approach consists of extracting the translation tuples from Wikipedia, disambiguating them by mapping them to WordNet word senses. We present results comparing two methods of disambiguation, the first using VSM on Wikipedia article texts and WordNet defi- nitions, and the second using categorical information extracted from Wikipedia, We find that mixing the two methods produces favorable results. Using the proposed method, we have constructed a multilingual Spanish-Japanese-English thesaurus con- sisting of 25,375 entries. The same method can be applied to any pair of languages that are linked to English in Wikipedia. 1 Introduction Aligned data resources are indispensable compo- nents of many Natural Language Processing (NLP) applications; however lack of annotated data is the main obstacle for achieving high performance NLP systems. Current success has been moderate. This is because for some languages there are few re- sources that are usable for NLP. Manual construction of resources is expensive and time consuming. For this reason NLP re- searchers have proposed semi-automatic or auto- matic methods for constructing resources such as dictionaries, thesauri, and ontologies, in order to facilitate NLP tasks such as word sense disam- biguation, machine translation and other tasks. Hoglan Jin and Kam-Fai Wong (2002) automati- cally construct a Chinese dictionary from different Chinese corpora, and Ahmad Khurshid et al. (2004) automatically develop a thesaurus for a specific domain by using text that is related to an image collection to aid in image retrieval. With the proliferation of the Internet and the immense amount of data available on it, a number of researchers have proposed using the World Wide Web as a large-scale corpus (Rigau et al., 2002). However due to the amount of redundant and ambiguous information on the web, we must find methods of extracting only the information that is useful for a given task. 1.1 Goals This research deals with the problem of developing a multilingual Japanese-English-Spanish thesaurus that will be useful to future Japanese-Spanish NLP research projects. A thesaurus generally means a list of words grouped by concepts; the resource that we create is similar because we group the words according to semantic relations. However, our resource is also 473 composed of three languages – Spanish, English, and Japanese. Thus we call the resource we created a multilingual thesaurus. Our long term goal is the construction of a Japa- nese-Spanish MT system. This thesaurus will be used for word alignments and building comparable corpus. We …

Content truncated to 500 words