Pronunciation Error Detection in Voice Input for Correct Word Suggestion
Kohei Kamimura, Kosuke Takano · 2019
In this paper, we present a method of pronunciation error detection in voice input for suggesting correct word. With the progress of the speech recognition technology, voice typing has been widely spread in many software and information services. However, to the best of our knowledge, there is no method to correct pronunciation errors for voice input by a user, although the speech recognition function can show only the words whose pronunciations are like the ones that the user input in voice. In this study, we propose a method for correcting pronunciation mistake in voice input by extending an existing method of spell mistake correction based on edit distance. By applying pronunciation score for phoneme sequences, which is proposed in our previous study, it is expected that our method can reduce the number of candidate pronunciation to be suggested in the process of pronunciation error detection.