Transcribers. In these 21 pairs, we count agreement pairs and disagreement pairs. Agreement is calculated as (1). For example, if a particular phone is labeled by six transcribers (T1-T6) as AH, but by one transcriber (T7) as AX, then the number of transcriber pairs who agree with each other is 15 (T1-T2~T6, T2-T3~6, T3-T4~6, T4-T5~6, T5-T6) and the number of transcriber pairs who disagree with each other is 6 (T1-T7, T2-T7, T3-T7, T4-T7, T5-T7, T6-T7). By (1), 15 pairs out of the 21 possible pair of transcriptions show percentage agreement as 0.714 (= 15/21).
Transcribers. The transcribers for each language were either CSLU sta or students. All were native speakers of the language they were labeling, or were considered very uent in the lan- guage. The English transcribers were VW (native) and TD (na- tive). They are professional phonetic transcribers, each of whom has completed a spectrogram reading course o ered by CSLU and has had extensive training and experience in labeling. The German transcribers were KB (native) and AJ (na- tive). KB has completed the spectrogram reading course, and XX received extensive on-the-job training. The Mandarin transcribers were LJ ( uent), YY ( uent), and ZH (native). Each completed the spectrogram reading course and received extensive training. Although there are three labelers, only two labeled each story. The Spanish transcribers were TL ( uent) and AJ (na- tive, see German). TL is trained in phonetics and has com- pleted the spectrogram reading course along with practical training.
Transcribers. The two transcribers were TL (also transcribed Spanish) and BO. Both are trained in phonetics, in spectrogram read- ing, and in the use of the OGI Speech tools.
Transcribers. The two labelers are trained in pho- netics and acoustics. Both are na- tive speakers of English and are fa- miliar with Spanish. They have less or no knowledge of the other lan- guages labeled. Both of the xxxx- scribers have had extensive experience labeling speech. DATA The data transcribed for this ex- periment were a subset of the OGI 22
Transcribers. 5.8.1. Although it is usual to engage the services of a transcriber in connection with a trial, arbitration, heavy or important application or deposition, prior approval must be obtained regarding the cost of such services, whether “live” or otherwise. Where reasonably possible at least two quotations must be secured, depending upon the level of service required (e.g., immediacy of transcription and speed of completion of the transcript.
