Levels of transcription in corpus linguistics-oriented research: the NERC proposal

 

Transcription of a spoken text for corpus-linguistic research can also be made at different levels. We will present here the proposal developed by French (1992) adopted by the NERC consortium. This system consists of four levels, numbered from one to four. Each successive level introduces more detail in the transcription, allowing several levels of detail according to different needs in particular types of research.

Level I

consists in the orthographic representation with minimal punctuation and without interactional information, so that change of speakers is not marked. The description of Level I includes conventions for orthographic representations and for punctuation.

Level II

is an enhanced orthographic representation with basic information about speaker identity, turn-taking, and non-verbal elements.

Level III

contains all the information included in Level II plus extra intonational and interactional information. Tone unit boundaries and tonic syllables are marked, precise identification of overlap onset and resolution are included. According to French (1992), transcription at this level needs to be done by trained phoneticians and a recording of substantial quality is necessary.

Level IV

is the most detailed level of transcription. It includes all information present in Level III plus additional intonational codings and acoustic and phonetic information. Tones, head syllables and a phonemic transcription are aligned with a digital representation of the wave form accompanied by a fundamental frequency tracing and a spectrogram of the utterances. A further possibility for this level would include tagging.