ISCA Archive Eurospeech 2001
ISCA Archive Eurospeech 2001

Reduction of alternative pronunciations in the norwegian computational lexicon norkompleks

Torbjørn Nordgård, Arne Kjell Foldvik

This paper describes a method for selecting one single pronunciation from a set of alternatives. Two types of reductions are described - base form reductions and inflected form reductions. The strategies are different because the inflected forms have to take care of two properties of Norwegian: The language allows for many spelling alternatives, both in base forms and inflected forms, and many morphological inflection types license pronunciation alternatives. A comprehensive pronunciation lexicon with more than 65000 base forms is used, and it is demonstrated that it is possible for users of the lexicon to derive their own stylistic preferences.