We would like to see Azure Speech-to-Text support in the next Bright Pattern release. We need it for both real-time conversations (chatbot support) and speech analytics.
According to our internal testing, Azure’s STT models achieve the best Word Error Rate (WER) for Baltic languages (Latvian, Lithuanian, Estonian) compared to current options.
Hi Alex,
Did you have a chance to test Bright Pattern STT?
As for Azure, is sounds like you require two different integrations - their real time STT and their batch STT, please confirm.
Sergey
Sergey,
We are currently testing BP STT, but so far we have limited information about its capabilities. What options are available for fine-tuning (for example, domain-specific or terminology-specific customization)? Is it possible to use corpus files, such as phrase lists or word vocabularies, during recognition?
Azure STT supports training with both text data (vocabularies) and audio data (transcriptions). This approach significantly improves recognition accuracy for domain-specific and even customer-specific data.
Regards,
Alex
