Background
Chatbots are becoming an essential feature of software applications. Its the new user interface. Microsoft supports Chatbots via Azure Bot Service. It supports multiple channels including text to speech and speech to text when hosted inside a web page using DirectLine channel.
Issue
We had to use Cognitive Services - Speech Services to get the text to speech and speech to text. The main is that when we started chatbot there was only Bing Speech integration. Then the progression was to Cognitive Services - Speech Services.
Around 06Feb2020 2PM EST, we could see that the speech service integration is not working in the development environment. We were troubleshooting some other issues and wasted some hours thinking the issue started because of our trial and error changes. Once we get back into the track, we tested in QA environment then further into Microsoft samples.
We could see the TTS and STT are not working there too.
We had Microsoft support, raised tickets and did escalations. They entered into the calls within hours. They started the normal procedure of asking for samples and all as we had some customizations in chatbot related to styling and all.
Workaround
During the process, we could see one sample that saying about web browser speech with BotFramework. It was added just one month back only. Link below
https://github.com/microsoft/BotFramework-WebChat/tree/master/samples/03.speech/f.web-browser-speech
It uses the speech capabilities provided by the Chrome browser which is part of W3C specifications. Since it is from browser, it should not have a dependency on Cognitive Services. As per caniuse.com Chrome supports the feature from early as v25
Luckily it worked. But the voice was not configurable like what we can do in Cognitive Services Speech service. Luckily we are only supporting Chrome as of now. This workaround should have been the solution. But another catch here is that the feature is still experimental as per Mozilla.
https://developer.mozilla.org/en-US/docs/Web/API/SpeechRecognition
https://developer.mozilla.org/en-US/docs/Web/API/SpeechRecognition
Solution
Ideally, the above workaround should be a solution for most of the teams which support only Chrome and its free from the browser. But since we work for the enterprise, the voice change was not acceptable. Money was not a problem to get the required voice so we just waited for Microsoft to fix. Microsoft fixed it within a couple of days.
Fortunately or unfortunately there is no entry into the Azure Status History and they have some wired explanations for it.
At last, everybody became happy.
The END
No comments:
Post a Comment