Cognates alignment

Cognates Alignment António Ribeiro1, Gaël Dias2, Gabriel Lopes1, João Mexia3 1Universidade Nova de Lisboa Faculdade de Ciências e Tecnologia Departamento de Informática Quinta da Torre, Monte da Caparica P–2829–516 Caparica Portugal {ambar,gpl}@di.fct.unl.pt 2Universidade da Beira Interior Unidade de Ciências Exactas Departamento de Informática Rua Marquês d’Ávila e Bolama P–6201–001 Covilhã Portugal ddg@noe.ubi.pt 3Universidade Nova de Lisboa Faculdade de Ciências e Tecnologia Departamento de Matemática Quinta da Torre, Monte da Caparica P–2829–516 Caparica Portugal Abstract Some authors (Simard et al.; Melamed; Danielsson & Mühlenbock) have suggested measures of similarity of words in different languages so as to find extra clues for alignment of parallel texts. Cognate words, like ‘Parliament’ and ‘Parlement’, in English and French respectively, provide extra anchors that help to improve the quality of the alignment. In this paper, we will extend an alignment algorithm proposed by Ribeiro et al. using typical contiguous and non-contiguous sequences of characters extracted using a statistically sound method (Dias et al.). With these typical sequences, we are able to find more reliable correspondence points and improve the alignment quality without recurring to heuristics to identify cognates. Keywords machine translation, alignment, cognates Introduction Alignment of parallel texts (texts which are mutual translations) is one of the first steps to be taken to build automatically a database of translation equivalents for bilingual lexicography or cross-lingual text processing tasks, such as machine(–aided) translation, cross-language information retrieval, multilingual question–answering systems to name but a few applications. Thus, it becomes crucial that those parallel texts should be as closely aligned as possible. That is to say, we should be able to make as detailed correspondences as possible between passages of texts and their translations in the other languages. Much work has already been done on sentence alignment, from early work by Brown et al. (1991), Gale & Church (1991) and Kay & Röscheisen (1993), to alignment of smaller text segments as in Simard et al. (1992), Church (1993), Fung & McKeown (1997), Melamed (1999) and Ribeiro et al. (2000a, b). Some methods have relied on using cognates, i.e. similar words like ‘Parliament’ and ‘Parlement’ (in English and French, respectively), in order to get extra clues for alignment. Several measures of “cognateness” have been suggested (Simard et al., 1992; Melamed 1999; Danielsson & Mühlenbock, 2000) but none is sufficiently reliable. That is, they do not provide any statistical studies supporting them and are tailored for specific applications. In this paper, we will extend a method of alignment proposed by Ribeiro et al. (2000a, b) by using typical contiguous and non-contiguous sequences of characters identified by statistical data analysis as shown in Dias et al. (2000b). We will start by giving an overview of several heuristics that have been proposed so far in order to identify cognates. In section 3, we will describe the methodology used to identify typical contiguous and non-contiguous sequences of characters and, in section 4, the alignment algorithm is presented. An evaluation of the results is given in section 5 and, finally, we will draw some …

Content truncated to 500 words