The Buckeye corpus of conversational speech: labeling conventions and a test of transcriber reliability

This paper describes the Buckeye corpus of spontaneous American English speech, a 307,000-word corpus containing the speech of 40 talkers from central Ohio, USA. The method used to elicit and record the speech is described, followed by a description of the protocol that was developed to phonemically label what talkers said. The results of a test of labeling consistency are then presented. The corpus will be made available to the scientific community when labeling is completed.