Jointly Trained Transformers models for Spoken Language Translation
원문 초록
Conventional spoken language translation (SLT) systems are pipeline based systems, where we have an Automatic Speech Recognition (ASR) system to convert the modality of source from speech to text and a Machine Translation (MT) systems to translate source text to text in target language. Recent progress in the sequence-sequence architectures have reduced the performance gap between the pipeline based SLT systems (cascaded ASR-MT) and End-to-End approaches. Though End-to-End and cascaded ASR-MT systems are reaching to the comparable levels of performances, we can see a large performance gap using the ASR hypothesis and oracle text w.r.t MT models. This performance gap indicates that the MT systems are prone to large performance degradation due to noisy ASR hypothesis as opposed to oracle text transcript. In this work this degradation in the performance is reduced by creating an end to-end differentiable pipeline between the ASR and MT systems. In this work, we train SLT systems with ASR objective as an auxiliary loss and both the networks are connected through the neural hidden representations. This train ing would have an End-to-End differentiable path w.r.t to the final objective function as well as utilize the ASR objective for better performance of the SLT systems. This architecture has improved from BLEU from 36.8 to 44.5. Due to the Multi-task training the model also generates the ASR hypothesis which are used by a pre-trained MT model. Combining the proposed systems with the MT model has increased the BLEU score by 1. All the experiments are reported on English-Portuguese speech translation task using How2 corpus. The final BLEU score is on-par with the best speech translation system on How2 dataset with no additional training data and language model and much less parameters.
함께 살펴볼 논문
공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.
Mlphon: A Multifunctional Grapheme-Phoneme Conversion Tool Using Finite State Transducers
TASK AWARE MULTI-TASK LEARNING FOR SPEECH TO TEXT TASKS
IWSLT 2026 인스트럭션 팔로잉 태스크를 위한 NAVER LABS 시스템 재구현
NAVER LABS System Re-implementation for the IWSLT 2026 Instruction-Following Task
우리는 NAVER LABS의 IWSLT 2025 인스트럭션 팔로잉 파이프라인을 IWSLT 2026 공유 태스크(제한 조건, 단음원 트랙)에 맞게 재구현하며, 지정된 구성 요소인 음성 인코더로서의 SeamlessM4T-v2-large와 대규모 언어 모델(LLM) 백본으로서의 Qwen3-4B-Instruct에 적응시킨다. 프로젝터 정렬, 텍스트 전용 LoRA 사전학습, 멀티모달 병합으로 이루어진 3단계 접근 방식은 원래 설계에서 그대로 유지된다. 또한 제공된 말뭉치로부터 열 가지 음성 중심 태스크 유형(태스크당 1만 개)에 걸쳐 총 10만 개의 합성 인스트럭션 팔로잉 예시를 구축하며, 이는 추가적인 3단계 미세조정에 적합하다. 우리의 주요 모델은 MCIF 벤치마크에서 EN-ZH 음성 번역에 대해 COMET 0.781을, 영어 SQA에 대해 BERTScore-F1 0.346을 달성한다.
Transformer Based Punctuation Restoration for Turkish
Explainability of Speech Recognition Transformers via Gradient-based Attention Visualization
Building a Public Domain Voice Database for Odia
토론
이 논문과 연결된 포럼 글이 아직 없습니다.