Intonation: Theory, Models, and Applications
This paper investigates the reliability of two prosodic labelling methods. We conducted experiments with 3 subjects on transcribing word prominence and prosodic boundary strength for American English. The reliability of transcription using two scales was compared: a scale of 4 values and a scale of unlimited values. We measured reliability in several ways, and obtained consistent results. The transcriptions yielded higher correlations using the smaller 4 value scale for prosodic boundary strength (PBS), but for word prominence the unlimited scale of values was applied more reliably. In practice the scale of unlimited values can be reduced to a scale of 10 values.
Bibliographic reference. Grover, C. / Heuft, Barbara / Coile, Bert Van (1997): "The reliability of labelling word prominence and prosodic boundary strength", In INT-1997, 165-168.