The Zero Resource Speech Benchmark 2021: Metrics and baselines for unsupervised spoken language modeling

TLDR

The paper introduces an unsupervised spoken language modeling task and the Zero Resource Speech Benchmark 2021, which evaluates models on phonetics, lexicon, syntax, and semantics without labels. The authors build a composite baseline that concatenates self‑supervised contrastive representation learning (CPC), k‑means clustering, and language modeling (LSTM or BERT) trained on pseudo‑text derived from the clustered representations. The pipeline achieves better‑than‑chance performance on all four metrics but falls short of text‑based topline systems, highlighting room for improvement.

Abstract

We introduce a new unsupervised task, spoken language modeling: the learning of linguistic representations from raw audio signals without any labels, along with the Zero Resource Speech Benchmark 2021: a suite of 4 black-box, zero-shot metrics probing for the quality of the learned models at 4 linguistic levels: phonetics, lexicon, syntax and semantics. We present the results and analyses of a composite baseline made of the concatenation of three unsupervised systems: self-supervised contrastive representation learning (CPC), clustering (k-means) and language modeling (LSTM or BERT). The language models learn on the basis of the pseudo-text derived from clustering the learned representations. This simple pipeline shows better than chance performance on all four metrics, demonstrating the feasibility of spoken language modeling from raw speech. It also yields worse performance compared to text-based 'topline' systems trained on the same data, delineating the space to be explored by more sophisticated end-to-end models.