The NIST Pilot Meeting Corpus has been entirely transcribed using a "quick" transcription procedure as detailed below. Meeting segments chosen for evaluation will be more carefully transcribed.

Quick Transcription:

The goal of quick transcription is to produce an accurate, time-aligned transcript as quickly and efficiently as possible to a level of quality suitable for system training (i.e., "quick" transcription). To this end a stripped-down transcription specification is required, which excludes special markup and multiple quality checks in favor of a single, focused transcription pass. The Quick Transcription procedure is available in PDF format.

The transcription data made by the LDC as an ECorpus (audio data order number LDC2004E01, transcriptions order number LDC2004E02) is available from the LDC, or directly from the tar-gzipped file LDC2004E02.tgz.