Resumo

CENDEJAS, Eduardo et al. Aligned Word Networks as a Resource for Extraction of Lexical Translation Equivalents, and their Application to the Text Alignment Task. Polibits [online]. 2012, n.46, pp.39-47. ISSN 1870-9044.

The notion of lexical translation equivalent is defined via correspondence established between two languages conventionally called source and target languages. We propose a method for extraction such equivalents for non-functional words. Our algorithm is based in two main resources: (1) MultiWordNet as a specialized lexicon for each one of the two languages in question and (2) a parallel text corpus as a source of additional information that provides various lexicalizations of the words that are being aligned. Our method is based on the fact that the word networks that form MultiWordNet are aligned. In addition, we discuss an application of the obtained list of word pairs in a word-level text alignment system. In our experiments we used bilingual sentence-level aligned parallel texts, without any morphosyntactic annotation, for word pairs Spanish / English and Spanish / Italian.

Palavras-chave : Translation equivalents; alignment; word networks; parallel texts.