UvT-WSD1: A Cross-Lingual Word Sense Disambiguation System

Proceedings of the 5th International Workshop on Semantic Evaluation, ACL 2010, pages 238–241, Uppsala, Sweden, 15-16 July 2010. c⃝2010 Association for Computational Linguistics UvT-WSD1: a Cross-Lingual Word Sense Disambiguation system Maarten van Gompel Tilburg centre for Cognition and Communication Tilburg University proycon@anaproy.nl Abstract This paper describes the Cross-Lingual Word Sense Disambiguation system UvT- WSD1, developed at Tilburg University, for participation in two SemEval-2 tasks: the Cross-Lingual Word Sense Disam- biguation task and the Cross-Lingual Lex- ical Substitution task. The UvT-WSD1 system makes use of k-nearest neighbour classifiers, in the form of single-word ex- perts for each target word to be disam- biguated. These classifiers can be con- structed using a variety of local and global context features, and these are mapped onto the translations, i.e. the senses, of the words. The system works for a given language-pair, either English-Dutch or English-Spanish in the current imple- mentation, and takes a word-aligned par- allel corpus as its input. 1 Introduction The UvT-WSD1 system described in this paper took part in two similar SemEval-2 tasks: Cross- Lingual Word Sense Disambiguation (Lefever and Hoste, 2010) and Cross-Lingual Lexical Substitu- tion (Mihalcea et al., 2010). In each task, a num- ber of words is selected for which the senses are to be determined for a number of instances of these words. For each word, a number of samples in context is provided, where each sample consists of one sentence, with the word to be disambiguated marked. Because of the cross-lingual nature of the tasks, a word sense corresponds to a translation in an- other language, rather than a sense description in the same language. In the Cross-lingual Lexical Substitution task, the target language is Spanish. The task is to find Spanish substitutes for the En- glish words marked in the test samples. In the Cross-Lingual Word Sense Disambiguation task, we participate for English-Dutch and English- Spanish. The Word Sense Disambiguation task provides training data for all five languages, in the form of the sentence-aligned EuroParl parallel cor- pus (Koehn, 2005). This is the source of training data the UvT-WSD1 system uses for both tasks. The system may output several senses per in- stance, rather than producing just one sense pre- diction. These are evaluated in two different ways. The scoring type “best” expects that the system outputs the best senses, in the order of its con- fidence. The scoring type “out of five/ten” ex- pects five or ten guesses, and each answer weighs the same. These metrics are more extensively described in (Mihalcea et al., 2010). The UvT- WSD1 system participates in both scoring types, for both tasks. The system put forth in this paper follows a similar approach as described in earlier research by (Hoste et al., 2002). 2 System Description The UvT-WSD1 system uses machine learning techniques to learn what senses/translations are as- sociated with any of the target words. It does so on the basis of a variety of local and global context features, discussed in Section 2.2. At the core of …

Content truncated to 500 words