The Zero Resource Speech Challenge 2015: Proposed Approaches and Results

By Maarten Versteegh, Xavier Anguera Miró, Aren Jansen, Emmanuel Dupoux. 2016

94 citations recorded by the source corpus.

This paper reports on the results of the Zero Resource Speech Challenge 2015, the first unified benchmark for zero resource speech technology, which aims at the unsupervised discovery of subword and word units from raw speech. This paper discusses the motivation for the challenge, its data sets, tasks and baseline systems. We outline the ideas behind the systems that were submitted for the two challenge tracks: unsupervised subword unit modeling and spoken term discovery, and summarize their results. The results obtained by participating teams show great promise; many systems beat the provided baselines and some even perform better than comparable supervised systems.

Subfields: Artificial Intelligence, Evaluation & Benchmarks, Speech & Audio

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Evaluation & Benchmarks

In Speech & Audio

Citations in the corpus

Libri-Light: A Benchmark for ASR with Limited or No Supervision (2019, positive): “The representation is evaluated using ABX error, a distance-based metric used in previous zero resource challenges [9, 10, 11].”
vq-wav2vec: Self-Supervised Learning of Discrete Speech Representations (2019, background): “Learning discrete representations of speech has gathered much recent interest (Versteegh et al., 2016; Dunbar et al., 2019).”
Unsupervised acoustic unit discovery for speech synthesis using discrete latent-variable neural networks (2019, neutral): “As part of the 2015 and 2017 Zero Resource Speech Challenges (ZRSC), several unsupervised methods were proposed for learning continuous acoustic representations [6–11].”
The zero resource speech challenge 2017 (2017, background): “hologists and clinicians interested in the impact of sociolinguisticvariationsininputonsubsequentnormalorabnormal language and cognitive development[9]. As the previous Zero Resource Speech Challenge [10, 11], the 2017 installment consists of two different subchallenges. Participants can submit syst”
The Zero Resource Speech Challenge 2019: TTS without T (2019, methodology): “6 Since it is unknown whether the discovered representations correspond to particular linguistic units (phone states, phonemes, features, syllables), we evaluate unit quality with a theory-neutral machine ABX score, as in previous Zero Resource challenges [1, 2].”
UWSpeech: Speech to Speech Translation for Unwritten Languages (2020, background): “r corresponding phonetic categories (discrete tokens) in an unsupervised manner, which mimics the way that human infants learn acoustic models in their mother tongue during their early years of life (Versteegh et al. 2016) (some of them only focus on a much easier task such as speech-totext translat”
A segmental framework for fully-unsupervised large-vocabulary speech recognition (2016, background): “Both human infants [39] and adults [40] use syllabic cues for word segmentation, and using such a bottom-up unsupervised syllabiﬁer can therefore be seen as one way to incorporate prior knowledge of the speech signal into a zero-resource system [41].”
Data Augmenting Contrastive Learning of Speech Representations in the Time Domain (2020, methodology): “This metric has been shown to be useful to analyse the linguistic content of speech features without having to train a classifier [23], and has been used in the Zero Resource Challenge series [9, 10, 24].”
An embedded segmental K-means model for unsupervised segmentation and clustering of speech (2017, methodology): “lustered into word-like units [17–21]. Two such full-coverage systems have recently been applied to the data of the Zero Resource Speech Challenge 2015 (ZRSC’15), giving a useful basis for comparison [22]. The ﬁrst is the Bayesian embedded segmental Gaussian mixture model (BES-GMM) [23]: a probabili”
Optimizing DPGMM Clustering in Zero Resource Setting Based on Functional Load (2018, methodology): “The ABX error rate is computed on the categories of the minimum pairs of triphones with distance measure as KL-divergence, as suggested by [1] when evaluating the posteriorgram representation.”
Early phonetic learning without phonetic categories: Insights from large-scale simulations on realistic input (2021, background): “training language, outperforming a number of alternatives that have been proposed for unsupervised speech representation learning (57, 58, 68).”
Virtual Phone Discovery for Speech Synthesis Without Text (2019, background): “Inspired from infant learning, zero resource spoken term discovery aims at discovering acoustic subword and word units from speech signals without the use of manual transcriptions or any linguistic knowledge [1, 2, 3].”
Visually Grounded Learning of Keyword Prediction from Untranscribed Speech (2017, background): “Here we consider the problem of grounding un-labelled speech when paired with images.”
Towards Unsupervised Learning of Speech Features in the Wild (2021, methodology): “, the Libri-light corpus [5], and most of the zero resource speech corpora [2, 3, 9]).”
Cyclic Spectral Modeling for Unsupervised Unit Discovery into Voice Conversion with Excitation and Waveform Modeling (2020, background): “Notably, in 2015 [4, 5], 2017 [7], and 2019 [10], the Zero Resource Speech Challenge (ZSC) has been organized, where the objective is to perform the discovery of subword/wordlike speech units (speech sounds representations) without any textual annotations, i.”
Unsupervised Speech Recognition via Segmental Empirical Output Distribution Matching (2018, methodology): “Unsupervised spoken term discovery Recently, the discovery of acoustic tokens including subword and word units has become a popular research topic (Dunbar et al. (2017); Versteegh et al. (2016); Burget et al.).”
ABX-Discriminability Measures and Applications. (Mesures de discriminabilité ABX et applications) (2016, background): “Results of the initial session of the challenge have been summarized in [98] and the challenge remains open online, with all the evaluation code and resources freely available.”