How will the machine distinguish different accents? How can the speech recognizer score content when pronunciation is very bad?
The speech recognizer has been trained and validated on samples of English from non-native speakers of English representing over 100 different native languages. Differences in ‘accent’ are thus accounted for in the acoustic models. This means that the representations of words in the system are designed to expect a wide variety of accented forms of English. The speech processor has been trained on over 126 different accents and can therefore deal with all of these accents equally. If the speaker has a very heavy accent due to non-native pronunciation such that would be assigned a low score by several human examiners, then this test taker will receive a low Pronunciation score from the machine also. ‘Manner of speech’ and ‘Content of speech’ are kept entirely distinct in the scoring. For example, if test takers use the correct vocabulary in their answer but they pronounce the word poorly, the speech processors will credit the Vocabulary score but debit the Pronunciation score accordingly
Related Questions
- How will the machine distinguish different accents? How can the speech recognizer score content when pronunciation is very bad?
- It is often very difficult to understand non-native speakers. How can a speech recognizer deal with different accents?
- How can the speech recognizer score content when pronunciation is very bad?
