It is often very difficult to understand non-native speakers. How can a speech recognizer deal with different accents?
For PTE Academic, close to half a million responses from more than ten thousand test takers originating from 158 home countries and speaking 126 different languages have been used to optimize the Pearson’s speech recognizer. Through this optimization process, Pearson has developed very advanced acoustic models which can understand even heavily accented speech from non-native speakers with poor pronunciation. No human rater is likely to be accustomed to over a hundred different foreign accents. So some human raters will have less difficulty with Indian speakers of English and where others feel more comfortable with Chinese speakers or French speakers. The Pearson automatic scoring system has been trained on 126 different accents and can therefore deal with all these accents equally.
Related Questions
- With so many different Spanish dialects and accents in Spanish, how do I determine which one to use so that everyone will understand?
- How will the machine distinguish different accents? How can the speech recognizer score content when pronunciation is very bad?
- How are different English accents evaluated by speakers? How do speakers see/judge different English accents?
