SENSEVAL-2 Japanese Translation Task

SENSEVAL-2 Japanese Translation Task Sadao K urohashi University of Tokyo kuro©kc.t.u-tokyo.ac.jp Abstract This paper reports an overview of SENSEVAL-2 Japanese translation task. In this task, word senses are defined according to translation dis- tinction. A translation Memory .· (TM) was constructed, which contains, for each Japanese head word, a list of typical Japanese expressions and their English translations. For each target word instance, a TM record best approximating that usage had to be submitted. Alternatively, submission could take the form of actual target word translations. 9 systems from 7 organiza- tions participated in the task. 1 Introduction In written texts, words which have multiple senses can be classified into two categories; homonyms and polysemous words. Generally speaking, while homonymy sense distinction is quite clear, polysemy sense distinction is very subtle and hard. English texts contain many homonyms. On the other hand, Japanese texts in which most content words are written by ideograms rarely contain homonyms. That is, the main target in Japanese WSD is polysemy, which makes Japanese WSD task setup very hard. What sense distinction of polysemous words is reasonable and effective heavily de- pends on how to use it, that is, an application ofWSD. Considering such a situation, in addition to the ordinary dictionary task we organized an- other task for Japanese, a translation task, in which word sense is defined according to trans- lation distinction. Here, we set up the task as- suming the example-based machine translation paradigm (Nagao, 1981). That is, first, a trans- lation memory (TM) is constructed which con- tains, for each Japanese head word, a list of typ- ical Japanese expressions (phrases/sentences) 37 involving the head word and an English trans- lation for each (Figure 1). We call a pair of Japanese and English expressions in the TM as a TM record. Given an evaluation document containing a target word, participants have to submit the TM record best approximating that usage. Alternatively, submissions can take the form of actual target word translations, or transla- tions of phrases or sentences including each tar- get word. This allows existing rule-based ma- chine translation (MT) systems to participate in the task, and we can compare TM based sys- tems with existing MT systems. For evaluation, we distributed newspaper ar- ticles. The number of target words was 40, and 30 instances of each target word were provided, making for a total of 1,200 instances. 2 Construction of Translation Memory The translation memory (TM) was constructed in two steps: 1. By referring to the KWIC (Key Word In Context) of a target word, its typical Japanese expressions are picked up by lex- icographers. 2. The Japanese expressions are translated by a translation company. KWIC was made from the nine years vol- ume of Mainichi Newspaper corpus. They are morphologically analyzed and segmented into phrase sequences, and then the 100 most fre- quent phrase uni-grams, hi-grams (two types; the target word is in the first phrase or the sec- ond phrase) and tri-grams (the target word is in the …

Content truncated to 500 words