COLT: a progress report

Phase 1 The COLT material was collected in London by a research team at the University of Bergen in 1993. It consists of roughly half-a-million words of spontaneous conversations between 13to 17-year old boys and girls from socially different school districts. During the period 1994–95, the conversations were transcribed orthographically (including indication of pauses and overlapping speech) by transcribers engaged by the Longman Group, and tagged for word-classes by a team at Lancaster university. In this form, COLT has become part of the British National Corpus (BNC). A demo of this version of COLT has been made available on Internet. At this point, the entire corpus has been checked and edited by the team in Bergen. The frequent occurrence of labels and the numerous instances of a question mark for speaker identity in the original transcripts indicate that the transcribers were faced with considerable problems. During our checking process, a great many instances of have disappeared, most of the speakers have been identified (with a substitution of the original names by fictitious ones), and mistakes in the original transcription have been straightened out. As a result, we have not only ended up with a transcription that is more faithful to the tape-recordings but also with a larger corpus; the number of words has increased by at least 15 per cent. This, in turn, has had the effect that the original word class tagging has become partly inadequate and that the edited corpus will have to be retagged. The retagging, which will be done by means of the Xanthippe software with assistance from Lancaster university, will be completed in the early ICAME Journal No. 20