TY - JOUR
T1 - Revisiting the effect of text complexity on Continuous Discourse Tracking using synthetic speech
T2 - Old tricks with new dogs
AU - Esmann Busch, Caroline
AU - Schaffalitzky de Muckadell, Charlotte
AU - Morris, David Jackson
PY - 2024
Y1 - 2024
N2 - Continuous Discourse Tracking (CDT) is a functional test of speech perceptual ability, which has been criticised on account of the procedural variation inherent in the method. This study sought to reduce this variation by using synthetic speech, which was subsequently vocoded to simulate listening with a cochlear implant. We also assessed the complexity of three text excerpts with auditory (n = 10) and written Cloze tests (n = 10). These same passages were used in an auditory-only CDT experiment (n = 12) performed with the synthetic- vocoded material. Mean tracking rates were lower, and the number of blockages was higher for the most difficult text as determined by the Cloze results. We also noted some anomalous realisations from the speech synthesis, but these were unlikely to have contributed to the differences in tracking rates that were observed for text complexity. These results show that Cloze testing is suitable to predict text complexity for CDT performed with synthesised speech. Furthermore, they indicate that the use of text-speech synthesis is viable and may be a useful addition to rehabilitation where functional measures are used to assess communication aptitude.
AB - Continuous Discourse Tracking (CDT) is a functional test of speech perceptual ability, which has been criticised on account of the procedural variation inherent in the method. This study sought to reduce this variation by using synthetic speech, which was subsequently vocoded to simulate listening with a cochlear implant. We also assessed the complexity of three text excerpts with auditory (n = 10) and written Cloze tests (n = 10). These same passages were used in an auditory-only CDT experiment (n = 12) performed with the synthetic- vocoded material. Mean tracking rates were lower, and the number of blockages was higher for the most difficult text as determined by the Cloze results. We also noted some anomalous realisations from the speech synthesis, but these were unlikely to have contributed to the differences in tracking rates that were observed for text complexity. These results show that Cloze testing is suitable to predict text complexity for CDT performed with synthesised speech. Furthermore, they indicate that the use of text-speech synthesis is viable and may be a useful addition to rehabilitation where functional measures are used to assess communication aptitude.
U2 - 10.1080/02699206.2023.2183104
DO - 10.1080/02699206.2023.2183104
M3 - Journal article
C2 - 36820623
VL - 38
SP - 172
EP - 183
JO - Clinical Linguistics & Phonetics
JF - Clinical Linguistics & Phonetics
SN - 0269-9206
IS - 2
ER -