Shashi Kumar, Andreas Stolcke, Malolan Chetlur, Bidisha Sharma, Karthik Durai, Shankar Venkatesan, Jeena Prakash, India Uniphore, Usa
We lifted 13 functions out of this paper's own repositories and ran 2 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.
| Repository | Role | Ran |
|---|---|---|
| k2-fsa/icefall | canonical | 0 of 11 |
| k2-fsa/sherpa | — | 2 of 2 |
| Function | Status | Where it lives |
|---|---|---|
| FrameReducer | Ran | k2-fsa/sherpa/triton/zipformer/model_repo_offline_bs/frame_reducer/1/model.py code served (permissive licence) · get_code("913ef6ad7c628305") |
| make_pad_mask | Ran | k2-fsa/sherpa/triton/zipformer/model_repo_offline_bs/frame_reducer/1/model.py code served (permissive licence) · get_code("3b57039980b0928c") |
| attach_diagnostics | Not yet run | k2-fsa/icefall/icefall/diagnostics.py code served (permissive licence) · get_code("4890d0a6d7925ab3") |
| average_checkpoints | Not yet run | k2-fsa/icefall/icefall/checkpoint.py code served (permissive licence) · get_code("c90ec0af48b591a2") |
| byte_decode | Not yet run | k2-fsa/icefall/icefall/byte_utils.py code served (permissive licence) · get_code("59d410aa19bdef96") |
| byte_encode | Not yet run | k2-fsa/icefall/icefall/byte_utils.py code served (permissive licence) · get_code("004eb0067bd43754") |
| convert_alignments_to_tensor | Not yet run | k2-fsa/icefall/icefall/ali.py code served (permissive licence) · get_code("8a64b592c82c0d10") |
| find_checkpoints | Not yet run | k2-fsa/icefall/icefall/checkpoint.py code served (permissive licence) · get_code("c75cf10a124dca6a") |
| get_class_name | Not yet run | k2-fsa/icefall/icefall/diagnostics.py code served (permissive licence) · get_code("b865558252e1566a") |
| get_tensor_stats | Not yet run | k2-fsa/icefall/icefall/diagnostics.py code served (permissive licence) · get_code("d6c6ed911b29d8a6") |
| load_alignments | Not yet run | k2-fsa/icefall/icefall/ali.py code served (permissive licence) · get_code("24af924ede3ca20e") |
| lookup_alignments | Not yet run | k2-fsa/icefall/icefall/ali.py code served (permissive licence) · get_code("a1f5678203ff47e8") |
| smart_byte_decode | Not yet run | k2-fsa/icefall/icefall/byte_utils.py code served (permissive licence) · get_code("226350299950338b") |
Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.
There has been increasing interest in unifying streaming and non-streaming automatic speech recognition (ASR) models to reduce development, training, and deployment costs. We present a unified framework that trains a single end-to-end ASR model for both streaming and non-streaming applications, leveraging future context information. We propose to use dynamic right-context through the chunked attention masking in the training of zipformer-based ASR models. We demonstrate that using rightcontext is more effective in zipformer models compared to other conformer models due to its multi-scale nature. We analyze the effect of varying the number of right-context frames on accuracy and latency of the streaming ASR models. We use Librispeech and large in-house conversational datasets to train different versions of streaming and non-streaming models and evaluate them in a production grade serverclient setup across diverse testsets of different domains. The proposed strategy reduces word error by relative 7.9% with a small degradation in user-perceived latency. By adding more right-context frames, we are able to achieve streaming performance close to that of nonstreaming models. Our approach also allows flexible control of the latency-accuracy tradeoff according to customers requirements.
The same record, over MCP at https://syntology.ai/mcp:
get_harvested_code_for_paper("2506.14434")
get_code_for_paper("2506.14434")
have("2506.14434")
Connect an agent — have() is free.