- Change Pykaldi with vosk-API (no python wrapper for decoding function, no extrat packages during installation, c++ implementation based on kaldi functions)
- New feature: Compute a confidence score per transcription
- Fix minor bugs
- Fix minor bugs
- put SWAGGER_PATH parameter as optional
- Generate the word_boundary file if it does not exist
- Speaker diarization feature: pyBK package
- Mulithreading feature: Speech decoding and Speaker diarization processes
- Optional parameter: real number of speaker in the audio
- Reimplement LinTO-Platform-stt-standalone-worker using Pykaldi package
- New features:
- Word timestamp computing
- Response type: plain/text: simple text output and application/json: the transcription and the words timestamp.
- Swagger: integrate swagger in the service using a python package
- Fix minor bugs
- First build of LinTO-Platform-stt-standalone-worker