paper-with-me

Papers

Label-Looping: Highly Efficient Decoding for Transducers

2024-06-10 · Vladimir Bataev, Hainan Xu, Daniel Galvez, Vitaly Lavrukhin, Boris Ginsburg

This paper introduces a highly efficient greedy decoding algorithm for Transducer-based speech recognition models. We redesign the standard nested-loop design for RNN-T decoding, swapping loops over frames and labels: the outer loop iterates over labels, while the inner loop iterates over frames searching for the next non-blank symbol. Additionally, we represent partial hypotheses in a special structure using CUDA tensors, supporting parallelized hypotheses manipulations. Experiments show that the label-looping algorithm is up to 2.0X faster than conventional batched decoding when using batch size 32. It can be further combined with other compiler or GPU call-related techniques to achieve even more speedup. Our algorithm is general-purpose and can work with both conventional Transducers and Token-and-Duration Transducers. We open-source our implementation to benefit the research community.

📄 PDF Abstract BibTeX arXiv:2406.06220

Code (1)

NVIDIA/NeMo 공식 구현 pytorch

Tasks

GPUspeech-recognitionSpeech Recognition

Similar Papers 제목 키워드 기반

Destination-Labeled Self-Looping Systems with Dwell: Intrinsic Characterization, Realization Cost, and Recognition

2026-06-29 · Reda Belaiche arxiv

We study a finite-state symbolic controller for systems in which the admissible visible transitions are fixed in advance and each visible state carries a minimum dwell requirement. The resulting model, which we call a de…

Integration of Frame- and Label-synchronous Beam Search for Streaming Encoder-decoder Speech Recognition

2023-07-24 · Emiru Tsunoo, Hayato Futami, Yosuke Kashiwagi, Siddhant Arora 외

Although frame-based models, such as CTC and transducers, have an affinity for streaming automatic speech recognition, their decoding uses no future knowledge, which could lead to incorrect pruning. Conversely, label-bas…

Automatic Speech RecognitionDecoderspeech-recognitionSpeech Recognition

Label-Synchronous Neural Transducer for End-to-End ASR

2023-07-06 · Keqi Deng, Philip C. Woodland

Neural transducers provide a natural way of streaming ASR. However, they augment output sequences with blank tokens which leads to challenges for domain adaptation using text data. This paper proposes a label-synchronous…

Domain AdaptationPrediction

Exact Decoding with Multi Bottom-Up Tree Transducers

2015-09-01 · WS 2015 9 · Daniel Quernheim
Language ModellingMachine Translation

VQ-T: RNN Transducers using Vector-Quantized Prediction Network States

2022-08-03 · Jiatong Shi, George Saon, David Haws, Shinji Watanabe 외

Beam search, which is the dominant ASR decoding algorithm for end-to-end models, generates tree-structured hypotheses. However, recent studies have shown that decoding with hypothesis merging can achieve a more efficient…

Language ModelingLanguage Modelling