A Multilingual Phoneme Inventory

Carole Tiberius (University of Brighton)


This paper explores the possibility of establishing a phoneme inventory 
for a group of related languages in which language-specific phonemes 
function as allophones of newly defined multilingual phonemes.

If we know, for example, that /A/ in Dutch is essentially equivalent to /{/ in English (as in "hand" /hAnt/ vs /h{nd/), we might be able to group these two language-specific phonemes into one multilingual phoneme, such as /{A/. (transcription is based on SAMPA)

The aim is to integrate such a multilingual phoneme inventory in a framework for Multilingual Lexical Representation which allows sharing of information between related languages at all levels of linguistic description (cf. Cahill and Gazdar 1999). The assumption is that a multilingual phoneme inventory will increase the amount of sharing at the phoneme and syllable level considerably.

This hypothesis has been tested on a subset of the Germanic languages, i.e. Dutch, English and, to a lesser degree, Danish and Icelandic.

In this paper, I will present the results of the cross-linguistic phoneme analysis. I will show how a phoneme inventory for a group of languages can be developed, and I will conclude by showing the advantages of integrating such a multilingual phoneme inventory in a framework for Multilingual Lexical Representation.