limitations
Assessing speech intelligibility is a critical component in the
assessment of communication efficacy2.
In
traditional evaluation the speaker remains the most influencing factor
because people could judge by ”feeling”.7-10 In
contrast, ASR system calculations are based on the input voice
information without being affected by other factors. Despite this
limitation, the results of the ASR system were highly consistent with
those of the traditional evaluation.
The limitation of ASR is that the evaluation score is too low when the
given speech is less intelligible. In most cases, the ASR score is lower
than the traditional assessment score. It does not mean that ASR
assessment does not show the true level of the child; on the contrary,
the ASR assessment is more sensitive to nonstandard pronunciation from
children because it can truly reflect the current speech level of the
children. However, it should be noted that different ASR software
programs and different versions might also affect the results, causing a
certain impact because of different internal dictionary systems.
Clinical applicability
We believe that the clinical intelligibility of children with hearing
loss can be evaluated using the ASR system. Further research is in
progress to enhance the possibilities of different versions of automatic
speech recognition, as it can also be of special interest as application
in a medical field. In addition, the ASR system can also try to assess
the articulation intelligibility of Mandarin-speaking children, such as
cleft lip and palate, dysarthria, childhood apraxia of speech and
aphasia. It can also further complete the intelligence of diagnosis and
directly match the recognition result with the correct answer to obtain
the final score.