2

I am using azure speech's fluency assessment. https://learn.microsoft.com/en-in/azure/cognitive-services/speech-service/how-to-pronunciation-assessment?pivots=programming-language-python

My questions is: the scoring returned keeping a native english speaker as reference. Is there a way to create score keeping a non-native english speaker in mind?

enter image description here

Anuj Gupta
  • 6,328
  • 7
  • 36
  • 55
  • 1
    Wouldn't that just be lower scores across the board? – Nick.Mc Mar 04 '21 at 06:45
  • @Nick.McDermaid not always there are those that are non-native but speak very good english. folks like them should not penalised – Anuj Gupta Mar 04 '21 at 08:17
  • 1
    Then they would get a good score? I'm probably missing the point here. – Nick.Mc Mar 04 '21 at 08:21
  • ideally yes but thats not what is happening! – Anuj Gupta Mar 05 '21 at 04:57
  • Currently Pronunciation assessment is only available for en-us & en-uk locales but I guess you can change the language locale to the variant of English say en-in (when it becomes available) instead of using the default en-us so that region specific pronunciations are considered - https://learn.microsoft.com/en-us/azure/cognitive-services/speech-service/language-support#neural-voices – mvark Dec 10 '21 at 07:55
  • This seems crazy. There are probably niches (acting, espionage) where acquiring an authentic non-native accent is important, but expecting a general-purpose tool to support this is just unrealistic and slightly weird. – tripleee May 06 '22 at 08:00

0 Answers0