Preliminary Recommendations

Development Phases

Three phases can be distinguished in the development of a corpus of spoken language:

1. The pre-recording phase: In this phase the objectives of the corpus are articulated, and from that the specification of the types of recording can be derived, including the physical situations, the number and type of speakers, the communicative events and the topic areas to be covered.
2. The recording phase: In this phase the management of the recording is devised and controlled -- the nature of any intervention, and the state of awareness of participants. The recording equipment is specified, and other technical details such as the placing of microphones.
3. the post-recording phase: In this phase the transcription is specified and carried out, and the transcription is processed for inclusion in the corpus

Note: In the assembly of data for the direct study of the speech signal, the above issues may well be replaced by others. The capture of genuine communicative events may not be important, whereas they are central for a general corpus of spoken language; the technical specifications may be much more elaborate; intervention and experiment are common, whereas they are excluded from material in the general corpus. The transcription and the analysis of the sound wave may be extremely detailed. The following types of speech data are regarded as special corpora from the perspective of general reference corpora, and are therefore not further considered here.

• read aloud isolated phonemes
• read aloud isolated words
• read aloud isolated sentences
• read aloud text fragments
• semi-spontaneous speech (e.g. numbers, alphanumerical expressions)
• spontaneous speech about a predetermined subject (e.g. retelling a story)
• factorial experiments (subjects performing in situations remote from normal communicative behaviour e.g. passing information about maps and routes (possibly across a visual barrier)).