| Téléchargement | - Voir la version finale : Towards effective and efficient non-autoregressive decoding using block-based attention mask (PDF, 2.5 Mio)
|
|---|
| DOI | Trouver le DOI : https://doi.org/10.21437/Interspeech.2024-404 |
|---|
| Auteur | Rechercher : Wang, Tianzi; Rechercher : Xie, Xurong; Rechercher : Li, Zhaoqing; Rechercher : Hu, Shoukang; Rechercher : Jin, Zengrui; Rechercher : Deng, Jiajun; Rechercher : Cui, Mingyu; Rechercher : Hu, Shujie; Rechercher : Geng, Mengzhe1; Rechercher : Li, Guinan; Rechercher : Meng, Helen; Rechercher : Liu, Xunying |
|---|
| Affiliation | - Conseil national de recherches Canada. Technologies numériques
|
|---|
| Format | Texte, Article |
|---|
| Conférence | Interspeech 2024, September 1 - 5, 2024, Kos, Greece |
|---|
| Sujet | speech recognition; non-autoregressive decoder; autoregressive decoder; bean search; conformer |
|---|
| Résumé | This paper proposes a novel non-autoregressive (NAR) block-based Attention Mask Decoder (AMD) that flexibly balances performance-efficiency trade-offs for Conformer ASR systems. AMD performs parallel NAR inference within contiguous blocks of output labels that are concealed using attention masks, while conducting left-to-right AR prediction and history context amalgamation between blocks. A beam search algorithm is designed to leverage a dynamic fusion of CTC, AR Decoder, and AMD probabilities. Experiments on the LibriSpeech-100hr corpus suggest the tripartite Decoder incorporating the AMD module produces a maximum decoding speed-up ratio of 1.73x over the baseline CTC+AR decoding, while incurring no statistically significant word error rate (WER) increase on the test sets. When operating with the same decoding real time factors, statistically significant WER reductions of up to 0.7% and 0.3% absolute (5.3% and 6.1% relative) were obtained over the CTC+AR baseline. |
|---|
| Date de publication | 2024-09-01 |
|---|
| Maison d’édition | ISCA |
|---|
| Dans | |
|---|
| Langue | anglais |
|---|
| Publications évaluées par des pairs | Oui |
|---|
| Exporter la notice | Exporter en format RIS |
|---|
| Signaler une correction | Signaler une correction (s'ouvre dans un nouvel onglet) |
|---|
| Identificateur de l’enregistrement | 9dad387d-f86d-4f4b-9d50-063ed037bd68 |
|---|
| Enregistrement créé | 2025-04-10 |
|---|
| Enregistrement modifié | 2025-04-15 |
|---|
