Anvendelse af korpora i opbygningen af ordbøger til maskinoversættelsessystemet METAL

Resumé

This paper presents some of the results that the Danish METAL group has achieved as to the production of lexicons for the MT system METAL, and especially our recommendations regarding the use of various documentation categories. Existing MT lexicons from the German-English language pair, lists ofterms for inhouse usage, as well as domain specific corpora form the basis ofthe lexicons. From experience we find that lexicons should mainly consist of words and terms extracted from technical corpora. However, technical documentation often consists of words that should not be entered into lexicons; this might be due to terminological inconsistence, the formation of ad hoc compounds, etc. I close this paper by dealing with some of the lexical problems we had to cope with when translating a specific technical text.