Draft October 2003 3 From recent overviews of annotated

This chapter describes the broad phonemic transcription in the CGN. First a broad overview of phonetic annotations in Dutch corpora is provided and a number of crucial dimensions are discussed: the source of annotation (human or automatic), the type of material involved, the level of transcription and the symbol set and transcription conventions. These dimensions serve as a guide through a number of aspects of the broad phonetic transcription in the CGN. In section 2 the level of transcription is discussed: methodological as well as fairly practical considerations are elaborated on with respect to the detail of phonetic annotations (or transcriptions) as well as the required level of expertise of the transcribers. In section 3 a pilot study is summarized that was meant to address issues such as the source of annotation and the transcription task (transcription ‘from scratch’ or verification of an automatically generated transcription) relative to practical matters such as the estimated transcription time, expected errors and variability. The next sections deal with the protocol: in section 4 the CGN set of phonetic symbols is defined and in section 5 the manual provided for the transcribers is described. The actual transcription procedure is dealt with in section 6: the entire corpus received an automatically generated broad phonetic transcription, ten percent of which, i.e. the ‘core corpus’, is manually verified. In section five the details of the grapheme-to-phoneme conversion of the orthographic annotation is described, as well as the manual verification procedure. The final section of this chapter is a first attempt to assess the quality of the manually verified transcriptions.