O n f e r e n c e

Results 
The mean ‘performance’ score in the present study for the 31 respondents over 20 
sentences was 4.56 which suggests a 91% accuracy rate. The corresponding mean 
‘opinion’ score was 3.2. The performance score is very comparable
to the results 
obtained by Al-Wabil, Al-Khalifa and Al-Saleh (2007). They reported a correctness level 
of 87% to 93% for non-vowelized Arabic words from the different TTS systems and from 
89% to 94% for vowelized words. While one cannot draw a direct comparison with the 
present study because of differences in the scoring methodologies, nevertheless, the 
similarity is informative. 
If one took these 
performance
scores alone then one would be tempted to conclude that 
the voices had reached a level of near perfection as is being claimed for the intelligibility 
of modern TTS synthesisers. One must, however, consider the 
opinion
score on the 
clarity of the speech so as to take into account the additional cognitive load associated 
with processing synthetic speech (van Merrienboer & Sweller, 2005). It is very much 
lower than the performance score and indicates that the listeners needed to give it 
greater attention than that which would be necessary for a human voice. 
In order to calculate an overall Intelligibility and Clarity Index the formula set out above 
was applied to the overall 20 sentences as follows: 
5 - 
∑
This formula was expanded to the format
5 - 
= 3.88 
The formula gives an overall value of 3.88, which is well below the perfect score of 5. The 
score indicates that the speech is considerably below the standard of natural speech. 
Since perfection eludes even the most advanced TTS systems this score seems quite 
positive. There are no similar values available from other studies since the combination 
of intelligibility and clarity has not been previously been considered in this manner.