The ELAN Slovene-English aligned corpus

MT Summit VII Sept. 1999 The ELAN Slovene-English Aligned Corpus Tomaž Erjavec Department for Intelligent Systems Jožef Stefan Institute Jamova 39 SI-1000 Ljubljana Slovenia Abstract Multilingual parallel corpora are a basic resource for research and development of MT. Such corpora are still scarce, espe- cially for lower-diffusion languages. The paper presents a sentence-aligned tokenised Slovene-English corpus, developed in the scope of the EU ELAN project. The cor- pus contains 1 million words from fifteen recent terminology-rich texts and is enco- ded according to the Guidelines for Text Encoding and Interchange (TEI). Our doc- ument type definition is a parametrisation of the TEI which directly encodes transla- tion units of the bi-texts. in a manner sim- ilar to that of translation memories. The corpus is aimed as a widely-distributable dataset for language engineering and for translation and terminology studies. The paper describes the compilation of the cor- pus, its composition, encoding and avail- ability. We highlight the corpus acquisi- tion and distribution bottlenecks and pre- sent our solutions. These have to do with the workflow in the project, and. not unre- latedly, with the encoding scheme for the corpus. 1 Introduction While translation systems are being developed inten- sely for major languages, there has been much less effort to secure MT or CAT tools for languages with a smaller number of speakers. Slovene is a South-Slavic language with ca. 2 million speakers and is spoken predominately in Slovenia. For the development as well as assessment of MT technology applied to a new language, it is extremely useful to have widely avail- able and reusable annotated corpora of the language in question. For MT research the most valuable type of corpus is one consisting of texts and their transla- tions, i.e., a parallel bilingual corpus. Further utility is obtained if the corpus is aligned at least at the sen- tence level, and the texts tokenised and part-of-speech annotated. The value of aligned corpora is well attested in practice, with several recent European projects de- voted to producing them, e.g., MLCC (Armstrong et al., 1998) (nine EU languages), Crater (McEnery et al., 1997) (Spanish, French and English), or ENPC (Johansson et al., 1996) (English-Norwegian). Such corpora have also been produced for non-European languages, e.g.. the HKUST Chinese-English corpus (Wu and Xia, 1995). For the Slovene language, the only available par- allel corpus has so far been the one released on the TELRI CD-ROM (Erjavec et al., 1998). which com- prises Plato's Republic and the MULTEXT-East cor- pus (Erjavec and Ide, 1998). The parallel part of the MULTEXT-East corpus consists of the novel '1984' by George Orwell in English and translations. The MUL- TEXT-East corpus derives most its value from the fact that it contains parallel texts in many languages, and is heavily annotated: the markup includes gross doc- ument structure, sentence and sub-sentence markup (names, quotes. ...), disambiguated lemmas and mor- phosyntactic descriptions of its words, and alignment of sentences with the ones from the English original. To facilitate the reusability of the …

Content truncated to 500 words