previous up contents next
Left: Multilingual lexical incompleteness Up: Multilingual lexicons for related Right: The DATR language

Summary

  We have proposed an architecture for multilingual lexical representation which aims to encode and exploit lexical similarities between closely related languages at all levels of linguistic description. We have shown how such an architecture can capture generalisations about English and German phonology, orthography and morphology, and we have noted that the same architecture can be used in a similar way to represent syntactic commonalities across languages.

There are several benefits to be gained from such an architecture for language pairs such as English/German over the traditional MT lexicon architecture. Our architecture provides a more economical encoding of lexical information and one which is easier to maintain. Just as inheritance lexicons generally provide the benefit of reduced redundancy in the lexicon and therefore more economical and transparent storage, so the use of inheritance across related languages will clearly provide the same benefits for multilingual lexicons.

Finally, there is the benefit of more intelligent approaches to lexical incompleteness. As discussed in the previous section, the architecture allows one to exploit default information from both source and target languages together with information about the default commonalities across those languages. The guesses will not always be right, but they will at least be intelligent guesses, guided by genuine linguistic principles, and in many cases, they will be close enough for comprehension by a native speaker.

---------------------------------------------------------

previous up contents next
Left: Multilingual lexical incompleteness Up: Multilingual lexicons for related Right: The DATR language
The PolyLex Web Pages. Copyright © Lynne Cahill & Gerald Gazdar, Tuesday 3 November 1998