IBM Watson对文本api的语音识别可以识别不同的重音吗？

Question

I am trying to use IBM Watson speech to text api. Does this api recognize different accents? Is there any way to teach is different accents, such as an Indian accent? In Windows, for speech recognition, accent recognition is done by providing some sample speech to understand the accent of the person. Is there a similar way in IBM Watson?

Answer 1

The IBM Watson Speech to Text API does not recognize accents by default. Depending on the language/accent, you'll have to look at using a specific language (full list: https://www.ibm.com/watson/developercloud/doc/speech-to-text/input.html#models ) and/or customizing your own model (available for only English and Japanese).

More info on customizing a model can be found here: https://www.ibm.com/watson/developercloud/doc/speech-to-text/custom.html

and an overview of the process from an IBM offering manager on the product is here: https://www.ibm.com/blogs/watson/2016/09/build-custom-language-model-convert-unique-speech-text/

These two resources should get you started, but as previously mentioned you'll have to test the effectiveness/performance for yourself.

Answer 2

IBM watson speech to text system is tailored to native speakers of American English. It does handle some forms of accented speech. You should try the service for your use case and see if the performance meets your needs.