DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Yang, Eunho | - |
dc.contributor.advisor | 양은호 | - |
dc.contributor.author | Song, Jaeyun | - |
dc.date.accessioned | 2022-04-27T19:32:02Z | - |
dc.date.available | 2022-04-27T19:32:02Z | - |
dc.date.issued | 2021 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=948454&flag=dissertation | en_US |
dc.identifier.uri | http://hdl.handle.net/10203/296130 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 전산학부, 2021.2,[iv, 25 p. :] | - |
dc.description.abstract | Despite the feature of real-time decoding, Monotonic Multihead Attention (MMA) shows comparable performance to the state-of-the-art offline methods in machine translation and automatic speech recognition (ASR) tasks. However, the latency of MMA is still a major issue in ASR and should be combined with a technique that can reduce the test latency at inference time, such as head-synchronous beam search decoding, which forces all non-activated heads to activate after a small fixed delay from the first head activation. In this paper, we remove the discrepancy between training and test phases by considering, in the training of MMA, the interactions across multiple heads that will occur in the test time. Specifically, we derive the expected alignments from monotonic attention by considering the boundaries of other heads and reflect them in the learning process. We validate our proposed method on the two standard benchmark datasets for ASR and show that our consistently trained version of MMA provides a better trade-off between quality and latency. | - |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | Online Speech Recognition▼aTransformer▼aMonotonic Multihead Attention▼aHead-Synchronous Beam Search Decoding▼aAutomatic Speech Recognition | - |
dc.subject | 실시간 음성인식▼a트랜스포머▼a모노토닉 멀티 헤드 어텐션▼a헤드-싱크로너스 빔 서치 디코딩▼a음성인식 | - |
dc.title | Monotonic multihead attention via mutually activating heads for online automatic speech recognition | - |
dc.title.alternative | 모노토닉 멀티헤드 어텐션의 헤드-싱크로너스 디코딩 학습을 통한 실시간 음성인식 기법 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 325007 | - |
dc.description.department | 한국과학기술원 :전산학부, | - |
dc.contributor.alternativeauthor | 송재윤 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.