Salvato in:
Dettagli Bibliografici
Autori principali: Huang, Hukai, Lu, Shenghui, Shan, Yahui, Qu, He, Zhang, Fengrun, Guan, Wenhao, Hong, Qingyang, Li, Lin
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2407.18581
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866916536867880960
author Huang, Hukai
Lu, Shenghui
Shan, Yahui
Qu, He
Zhang, Fengrun
Guan, Wenhao
Hong, Qingyang
Li, Lin
author_facet Huang, Hukai
Lu, Shenghui
Shan, Yahui
Qu, He
Zhang, Fengrun
Guan, Wenhao
Hong, Qingyang
Li, Lin
contents The Mixture of Experts (MoE) model is a promising approach for handling code-switching speech recognition (CS-ASR) tasks. However, the existing CS-ASR work on MoE has yet to leverage the advantages of MoE's parameter scaling ability fully. This work proposes DLG-MoE, a Dynamic Language Group-based MoE, which can effectively handle the CS-ASR task and leverage the advantages of parameter scaling. DLG-MoE operates based on a hierarchical routing mechanism. First, the language router explicitly models the language attribute and dispatches the representations to the corresponding language expert groups. Subsequently, the unsupervised router within each language group implicitly models attributes beyond language and coordinates expert routing and collaboration. DLG-MoE outperforms the existing MoE methods on CS-ASR tasks while demonstrating great flexibility. It supports different top-$k$ inference and streaming capabilities and can also prune the model parameters flexibly to obtain a monolingual sub-model. The code has been released.
format Preprint
id arxiv_https___arxiv_org_abs_2407_18581
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Dynamic Language Group-Based MoE: Enhancing Code-Switching Speech Recognition with Hierarchical Routing
Huang, Hukai
Lu, Shenghui
Shan, Yahui
Qu, He
Zhang, Fengrun
Guan, Wenhao
Hong, Qingyang
Li, Lin
Computation and Language
Artificial Intelligence
The Mixture of Experts (MoE) model is a promising approach for handling code-switching speech recognition (CS-ASR) tasks. However, the existing CS-ASR work on MoE has yet to leverage the advantages of MoE's parameter scaling ability fully. This work proposes DLG-MoE, a Dynamic Language Group-based MoE, which can effectively handle the CS-ASR task and leverage the advantages of parameter scaling. DLG-MoE operates based on a hierarchical routing mechanism. First, the language router explicitly models the language attribute and dispatches the representations to the corresponding language expert groups. Subsequently, the unsupervised router within each language group implicitly models attributes beyond language and coordinates expert routing and collaboration. DLG-MoE outperforms the existing MoE methods on CS-ASR tasks while demonstrating great flexibility. It supports different top-$k$ inference and streaming capabilities and can also prune the model parameters flexibly to obtain a monolingual sub-model. The code has been released.
title Dynamic Language Group-Based MoE: Enhancing Code-Switching Speech Recognition with Hierarchical Routing
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2407.18581