Klaske van Sluis

86 5.3. Results 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 Averaged rater responses per speech fragment Individual rater response 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 22 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 33 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 5 5 5 5 5 55 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 77 7 77 7 7 7 7 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 88 8 8 8 8 8 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 99 9 9 99 9 9 9 9 0 200 400 600 800 1000 0 200 400 600 800 1000 Intelligibility 0−9: SLPs R^2 = 0.51 0 00 0 0 00 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 Averaged rater responses per speech fragment Individual rater response 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 11 2 2 2 2 2 222 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 3 33 3 3 3 3 3 3 3 3 3 3 3 3 3 33 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 44 4 4 4 4 4 4 4 4 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 55 5 5 5 5 5 5 5 5 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 7 7 7 7 7 77 7 7 7 7 7 77 7 7 7 7 7 7 7 7 7 7 7 7 8 88 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 0 200 400 600 800 1000 0 200 400 600 800 1000 Voice Quality 0−9: SLPs R²=0.47 Asisto: AVQI R²=0.37 (p < 0.002) Figure 5.1: Variation in perceptual evaluations and automatic scores in experiment 1. Numbers: individual expert raters; filled symbols: automatic ASISTO scores. Cor- relations are with average expert responses. thus also to intelligibility ( R =0.87 and R =0.84, p ≤ .001). This makes AVQI a better predictor of perceptual intelligibility in experiment 2 than the ELIS scores. Differences between periods I and II in Figure 5.2 were not significant for ELIS or AVQI ( p > .025, Bonferroni correction). Overall, five speakers had statistically significant worse intelligibility at T2 (T2-T1 < 0), two speakers were better at T2 (T2-T1 > 0), and four were neither better nor worse (T2-T1 0), see Figure 5.2. One speaker was scored with worse intelligibility at T2 and unchanged voice quality (WWL). In total, there are roughly as many speakers that showed a decline in intelligibility and voice quality at T2 as showed unchanged or improved intelligibility and voice quality. The ELIS scores tended to score the T2 as more intelligible than the T1 recordings. Currently, it is not clear how to interpret this difference with perceptual intelligibility scores. The AVQI scores were distributed more like the corresponding perceptual voice quality scores, in line with the high correlation between AVQI and voice quality scores. In this sample of 13 speakers, three distinct levels of change can be distinguished: better at T2, worse at T2, and no difference. When speakers from each of these levels are compared against speakers from other levels a statistical significant difference is found (p < .001). Together, the automatic and perceptual results presented in Figure 5.2 indicate that there is no definite trend in the changes in intelligibility and voice quality after 7 years or more. There might be a somewhat bigger probability for a decline in intelligibility and voice quality than the reverse. However, it is clear that the differences between speakers in direction and extend of change over time are large.