SENS-ASRM
Date
Formats
Categories
Content
SENS-ASRM is a research project dedicated to improving streaming automatic speech recognition under low-latency constraints.
The method enriches acoustic representations with semantic information inferred from past audio context, helping transcription models remain more accurate when future context is unavailable.
By making the implementation publicly available, Orange supports reproducibility, fosters reuse by the research and developer communities, and helps accelerate innovation in real-time speech technologies.
The project is available on GitHub under the MIT licence.