Multilingual Lexicons for Related Languages

The great increase in work on the lexicon by computational and theoretical linguists throughout the 1980s has concerned itself, almost exclusively, with monolingual lexicons. Meanwhile, applied work on multilingual lexicons, mostly for machine translation (MT), has employed monolingual lexicons linked only at the level of semantics. In this paper, we argue that the traditional MT lexicon architecture, while arguably adequate for unrelated languages, makes it impossible to capture important generalisations about related languages. These generalisations, if captured, can help to produce more robust multilingual natural language processing (NLP) systems for such languages. We discuss the types of relation we find in Dutch, English and German and describe a hierarchical lexicon structure which permits sharing of information across any level of lexical representation, illustrating it with examples from English and German. 1 Introduction In contrast to linguistic work of the 1960s and 1970s..