SpeechDPR: End-to-End Spoken Passage Retrieval for Open-Domain Spoken Question Answering
Chyi-Jiunn Lin, Guan-Ting Lin, Yung-Sung Chuang, Wei-Lun Wu, Shang-Wen Li, Abdelrahman Mohamed, Hung-yi Lee, Lin-shan Lee
TL;DR
SpeechDPR tackles openSQA by providing an end-to-end spoken passage retriever that operates without supervised ASR or speech-text pairs. It distills knowledge from a cascading teacher (UASR + TDR) into a bi-encoder model that encodes spoken questions and passages into a shared semantic space, with a training objective that combines end-to-end NLL losses and teacher guidance. Empirical results show SpeechDPR achieves retrieval performance comparable to cascading baselines and exhibits superior robustness as ASR quality deteriorates, while ensemble combinations can surpass the baselines. This approach enables robust openSQA in scenarios with limited labeled data and supports future extension to low-resource languages where paired speech-text data are scarce.
Abstract
Spoken Question Answering (SQA) is essential for machines to reply to user's question by finding the answer span within a given spoken passage. SQA has been previously achieved without ASR to avoid recognition errors and Out-of-Vocabulary (OOV) problems. However, the real-world problem of Open-domain SQA (openSQA), in which the machine needs to first retrieve passages that possibly contain the answer from a spoken archive in addition, was never considered. This paper proposes the first known end-to-end framework, Speech Dense Passage Retriever (SpeechDPR), for the retrieval component of the openSQA problem. SpeechDPR learns a sentence-level semantic representation by distilling knowledge from the cascading model of unsupervised ASR (UASR) and text dense retriever (TDR). No manually transcribed speech data is needed. Initial experiments showed performance comparable to the cascading model of UASR and TDR, and significantly better when UASR was poor, verifying this approach is more robust to speech recognition errors.