Papers by Shashi Kumar
Unifying Streaming and Non-streaming Zipformer-based ASR (2025.acl-industry)
Copied to clipboard
Bidisha Sharma, Karthik Pandia D S, Shankar Venkatesan, Jeena J Prakash, Shashi Kumar, Malolan Chetlur, Andreas Stolcke
| Challenge: | Existing frameworks for streaming and non-streaming ASR models have been used to reduce development, training and deployment costs. |
| Approach: | They propose to use dynamic right-context through chunked attention masking to train zipformer-based ASR models. |
| Outcome: | The proposed framework reduces word error by relative 7.9% with a small degradation in user-perceived latency. |
TokenVerse: Towards Unifying Speech and NLP Tasks via Transducer-based ASR (2024.emnlp-main)
Copied to clipboard
Shashi Kumar, Srikanth Madikeri, Juan Pablo Zuluaga Gomez, Iuliia Thorbecke, Esaú Villatoro-tello, Sergio Burdisso, Petr Motlicek, Karthik S, Aravind Ganapathiraju
| Challenge: | Existing approaches to automatic speech recognition use cascaded pipelines for tasks like voice activity detection, diarization, transcription and subsequent processing. |
| Approach: | They propose a single Transducer-based model that integrates task-specific tokens into the reference text during ASR model training, streamlining inference and eliminating the need for separate NLP models. |
| Outcome: | The proposed model outperforms the existing pipeline on speaker change detection, endpointing, and NER tasks while outperforming the existing model in individual task performance. |
Fast Streaming Transducer ASR Prototyping via Knowledge Distillation with Whisper (2024.findings-emnlp)
Copied to clipboard
Iuliia Thorbecke, Juan Pablo Zuluaga Gomez, Esaú Villatoro-tello, Shashi Kumar, Pradeep Rangappa, Sergio Burdisso, Petr Motlicek, Karthik S, Aravind Ganapathiraju
| Challenge: | a recent study shows that training of ASR models with little to no supervised data is challenging. |
| Approach: | They propose a framework to train streaming Transformer-Transducer models with pseudo-labeled (PL) speech from foundational speech models. |
| Outcome: | The proposed framework can be trained from scratch with pseudo-labeled speech from foundational speech models (FSMs) the proposed framework is validated on 6 languages from CommonVoice and proposes multiple heuristics to filter out hallucinated PLs. |