An embedded segmental K-means model for unsupervised segmentation and clustering of speech
This addresses the problem of efficient zero-resource speech processing for researchers, offering a faster alternative to Bayesian methods with competitive performance, though it is incremental as it approximates an existing model.
The paper tackled unsupervised segmentation and clustering of unlabeled speech by introducing an embedded segmental K-means model (ES-KMeans), which outperformed a leading heuristic method in word segmentation with similar scores to a Bayesian model while being 5 times faster and scaling to larger corpora up to 45 hours.
Unsupervised segmentation and clustering of unlabelled speech are core problems in zero-resource speech processing. Most approaches lie at methodological extremes: some use probabilistic Bayesian models with convergence guarantees, while others opt for more efficient heuristic techniques. Despite competitive performance in previous work, the full Bayesian approach is difficult to scale to large speech corpora. We introduce an approximation to a recent Bayesian model that still has a clear objective function but improves efficiency by using hard clustering and segmentation rather than full Bayesian inference. Like its Bayesian counterpart, this embedded segmental K-means model (ES-KMeans) represents arbitrary-length word segments as fixed-dimensional acoustic word embeddings. We first compare ES-KMeans to previous approaches on common English and Xitsonga data sets (5 and 2.5 hours of speech): ES-KMeans outperforms a leading heuristic method in word segmentation, giving similar scores to the Bayesian model while being 5 times faster with fewer hyperparameters. However, its clusters are less pure than those of the other models. We then show that ES-KMeans scales to larger corpora by applying it to the 5 languages of the Zero Resource Speech Challenge 2017 (up to 45 hours), where it performs competitively compared to the challenge baseline.