Computer-coded phonetic transcription

Abstract

Speech databases on Danish, Dutch, English, French, German and Italian are being set up as part of an Esprit-funded project, ‘Speech Assessment Methods’ (SAM), which is concerned with the establishment of standardized assessment methods for automatic speech recognition devices. An important aspect of this work is the devising of a standard set of conventions for the machine-readable phonetic transcription of these languages. This matter was discussed at a two-day meeting in London devoted to aspects of the organization of speech databases, which was attended by participants from SAM Partners in the five currently participating countries (Denmark, France, Italy, Netherlands, United Kingdom), together with representatives from the German Federal Republic. Participants agreed that a compromise based on the three existing multilingual proposals could be achieved: its broad outlines were agreed, while the elaboration of detail was left to the present author and this report.