Unsupervised word segmentation and lexicon discovery using acoustic word embeddings
This addresses the challenge of developing speech technology without transcriptions, which is crucial for low-resource languages or modeling infant language acquisition, though it is incremental as it builds on prior unsupervised methods.
The authors tackled the problem of unsupervised word segmentation and lexicon discovery from unlabelled speech data, achieving around 20% word error rate in a connected digit recognition task, which outperformed a previous HMM-based system by about 10% absolute.
In settings where only unlabelled speech data is available, speech technology needs to be developed without transcriptions, pronunciation dictionaries, or language modelling text. A similar problem is faced when modelling infant language acquisition. In these cases, categorical linguistic structure needs to be discovered directly from speech audio. We present a novel unsupervised Bayesian model that segments unlabelled speech and clusters the segments into hypothesized word groupings. The result is a complete unsupervised tokenization of the input speech in terms of discovered word types. In our approach, a potential word segment (of arbitrary length) is embedded in a fixed-dimensional acoustic vector space. The model, implemented as a Gibbs sampler, then builds a whole-word acoustic model in this space while jointly performing segmentation. We report word error rates in a small-vocabulary connected digit recognition task by mapping the unsupervised decoded output to ground truth transcriptions. The model achieves around 20% error rate, outperforming a previous HMM-based system by about 10% absolute. Moreover, in contrast to the baseline, our model does not require a pre-specified vocabulary size.