https://scholars.lib.ntu.edu.tw/handle/123456789/580909
標題: | Improving automatic speech recognition and speech translation via word embedding prediction | 作者: | Chuang S.-P Liu A.H Sung T.-W Lee H.-Y. HUNG-YI LEE |
關鍵字: | Audio signal processing; Embeddings; Semantics; Speech; Speech communication; Automatic speech recognition; Cascaded system; Contextual information; Decoding methods; Intermediate representations; Semantic relations; Speech translation; Spoken language processing; Speech recognition | 公開日期: | 2021 | 卷: | 29 | 起(迄)頁: | 93-105 | 來源出版物: | IEEE/ACM Transactions on Audio Speech and Language Processing | 摘要: | In this article, we target speech translation (ST). We propose lightweight approaches that generally improve either ASR or end-to-end ST models. We leverage continuous representations of words, known as word embeddings, to improve ASR in cascaded systems as well as end-to-end ST models. The benefit of using word embedding is that word embedding can be obtained easily by training on pure textual data, which alleviates data scarcity issue. Also, word embedding provides additional contextual information to speech models. We motivate to distill the knowledge from word embedding into speech models. In ASR, we use word embeddings as a regularizer to reduce the WER, and further propose a novel decoding method to fuse the semantic relations among words for further improvement. In the end-to-end ST model, we propose leveraging word embeddings as an intermediate representation to enhance translation performance. Our analysis shows that it is possible to map speech signals to semantic space, which motivates future work on applying the proposed methods in spoken language processing tasks. ? 2014 IEEE. |
URI: | https://www.scopus.com/inward/record.uri?eid=2-s2.0-85097717427&doi=10.1109%2fTASLP.2020.3037543&partnerID=40&md5=2c7afb0791f8e804b59e3df020443879 https://scholars.lib.ntu.edu.tw/handle/123456789/580909 |
ISSN: | 23299290 | DOI: | 10.1109/TASLP.2020.3037543 |
顯示於: | 電機工程學系 |
在 IR 系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。