Do We Still Need Automatic Speech Recognition for Spoken Language Understanding?
This addresses the efficiency and accuracy of SLU systems for applications like emergency response and named entity recognition, though it is incremental as it builds on existing representation learning methods.
The study investigated whether learned speech features from wav2vec 2.0 could replace automatic speech recognition (ASR) transcripts in spoken language understanding (SLU) tasks, finding that speech features outperformed ASR on three classification tasks but ASR was better for machine translation.
Spoken language understanding (SLU) tasks are usually solved by first transcribing an utterance with automatic speech recognition (ASR) and then feeding the output to a text-based model. Recent advances in self-supervised representation learning for speech data have focused on improving the ASR component. We investigate whether representation learning for speech has matured enough to replace ASR in SLU. We compare learned speech features from wav2vec 2.0, state-of-the-art ASR transcripts, and the ground truth text as input for a novel speech-based named entity recognition task, a cardiac arrest detection task on real-world emergency calls and two existing SLU benchmarks. We show that learned speech features are superior to ASR transcripts on three classification tasks. For machine translation, ASR transcripts are still the better choice. We highlight the intrinsic robustness of wav2vec 2.0 representations to out-of-vocabulary words as key to better performance.