Sources of linguistic knowledge for minority languages

MT Summit VII Sept. 1999 Sources of Linguistic Knowledge for Minority Languages Harold L. Somers Department of Language Engineering UMIST, Manchester, England harold@ccl.umist.ac.uk Abstract Language Engineering (LE) products and resources for the world’s “major” languages are steadily increasing, but there remains a major gap as regards less widely-used languages. This paper considers the current situation regarding LE resources for some of the languages in question, and some proposals for rectifying this situation are made, including techniques based on adapting existing resources and “knowledge extraction” techniques from machine- readable corpora. Key words: minority languages, corpora, knowledge extraction 1. Introduction While the availability of Language Engineering (LE) products and resources for the world’s “major” languages steadily increases, including Machine Translation systems, CAT systems, on-line dictionaries, thesauri, and so on, there remains a major gap as regards less widely-used languages.1 Missing are not only these kinds of products, but even simple tools like spelling- and grammar- checkers. Because of accidents of world politics as much as anything else, the world's languages fall into three or four ranks, reflecting the computational resources available for them. This paper will identify some languages which are more or less badly served, and will make some proposals for what we can do about the situation. Recognising that the development of LE products for a new language is rarely a trivial matter, we will suggest some 1 An earlier version of this paper appeared in the proceedings of Translating and the Computer 19 (Aslib). techniques that can make the task more manageable, or more feasible. 2. Minority Languages The notion of “minority language” is relative, depending on the geographical stand-point of the observer. We can define the term from an LE perspective (see below), or else from a local point of view. This latter option is relevant, since our proposed solution to the problem of linguistic knowledge acquisition relies on there being a community of professional linguists servicing the minority language speaking community. The UK is nominally an English-speaking country, with small regions where the indigenous Celtic languages are more or less widely spoken. However, a more realistic linguistic profile of the UK must take into account that there are significant groups of people speaking non-indigenous minority languages (NIMLs). Across the country, languages from the Indian subcontinent, as well as Cantonese, are widely spoken; other NIMLs are more regionally concentrated, e.g. Greek and Turkish in London. In other countries, the picture will be different, but only in the details. While second- and third-generation immigrants are largely proficient in English, having received their schooling in this country, new immigrants as well as older members of the immigrant communities — especially women — are often functionally illiterate in English, even if they are long-term residents (Rudat, 1994). Many local councils, particularly in urban areas, recognize this, and maintain language departments to provide translation and interpreting services with in-house staff as well as lists of freelance translators. Their work includes translating information leaflets about - 531- MT Summit VII Sept. 1999 community …

Content truncated to 500 words