823-OLT @ BUET DL Sprint 4.0: Context-Aware Windowing for ASR and Fine-Tuned Speaker Diarization in Bengali Long Form Audio
1 st Ratnajit Dhar, 2 nd Arpita Mallik · 2026
Read 823-OLT @ BUET DL Sprint 4.0: Context-Aware Windowing for ASR and Fine-Tuned Speaker Diarization in Bengali Long Form Audio by 1 st Ratnajit Dhar, 2 nd Arpita Mallik (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-21183.txt.
Description
Bengali is widely spoken yet poorly served by long-form speech technology. Ratnajit Dhar and Arpita Mallik present frameworks for Bengali long-form speech intelligence in this 2026 arXiv paper, covering transcription and speaker attribution. Speech recognition uses a Whisper-based model with vocal separation, voice activity detection, and gap-aware windowing that preserves context for stable decoding.
Speaker diarization fine-tunes a pretrained segmentation model on the official competition dataset to capture Bengali conversational patterns. Together the systems deliver efficient transcription plus speaker-aware transcripts for long recordings. The work offers scalable speech technology for a low-resource language, built for a competition setting and beyond.
Catalog record
| CREATOR | 1 st Ratnajit Dhar, 2 nd Arpita Mallik |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 18 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-21183