Evaluating speech features with the minimal-pair ABX task (II): resistance to noise

By Thomas Schatz, Vijayaditya Peddinti, Xuan-Nga Cao, Francis R. Bach, Hynek Hermansky, Emmanuel Dupoux. 2014

38 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Signal Processing

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 Computer Vision and Pattern Recognition

In Signal Processing

Citations in the corpus

The zero resource speech challenge 2017 (2017, methodology): “We use a minimal pair ABX task [18, 19], which does not require any training, and only requires a dissimilarity metric between speech tokens.”
The Zero Resource Speech Challenge 2015: Proposed Approaches and Results (2016, methodology): “For this reason, in the present challenge, we took a different approach and evaluated phoneme discriminability directly on the feature representation using the Minimal-Pair ABX (MP-ABX) task [15, 16].”
Parallel inference of dirichlet process Gaussian mixture models for unsupervised acoustic modeling: a feasibility study (2015, methodology): “Following the track 1 of the challenge, our evaluation metric is error rate in the ABX discriminability task [26, 27].”
A deep scattering spectrum — Deep Siamese network pipeline for unsupervised acoustic modeling (2016, methodology): “This evaluation is done by computing an ABX score [8, 9].”
Learning Weakly Supervised Multimodal Phoneme Embeddings (2017, methodology): “To evaluate the phonetic discriminability of our embeddings, we use an ABX discrimination task [24, 25].”
Linguistic Unit Discovery from Multi-Modal Inputs in Unwritten Languages: Summary of the “Speaking Rosetta” JSALT 2017 Workshop (2018, background): “2 [21]) and the ABX task (which compares the similarity between discovered units and ground truth labels or between different types of acoustic features; [22, 23]) for the evaluation of the discovered units, and quantitative measures, such as BLEU score, error rates, and word discovery metrics (see ”
Deep learning methods for unsupervised acoustic modeling — Leap submission to ZeroSpeech challenge 2017 (2017, methodology): “Recently, several approaches have been proposed for this task using various feature representations [7] and neural network models [8, 9].”
Multilingual bottle-neck feature learning from untranscribed speech (2017, methodology): “For further detailed definition, please refer to [23, 24] or the website of ZeroSpeech2017.”
A comparison of neural network methods for unsupervised representation learning on the zero resource speech challenge (2015, unknown): “Following the ZRSC’s protocol, we evaluate our frame representations using an ABX task [24, 25] which measures the discriminability of frame representations by asking whether triphone x is most like triphone a or triphone b, where a and x are distinct examples of the same triphone sequence and b is ”
A Brief Overview of Unsupervised Neural Speech Representation Learning (2022, background): “The longstanding zero resource speech challenge (ZeroSpeech) offers a new set of tasks for each edition (Versteegh et al. 2015; Dunbar et al. 2017, 2019, 2020, 2021) usually featuring a minimal-pair ABX task (Schatz et al. 2013, 2014).”
ABX-Discriminability Measures and Applications. (Mesures de discriminabilité ABX et applications) (2016, background): “They showed how the performance of such systems can be affected by the choice of acoustic features [35, 88, 96, 103], of an acoustic modeling method [87, 90, 97, 102, 105, 106] or the combination of both [107, 108].”
The zero resource speech challenge 2015 (2015, methodology): “Here, we will use a minimal pair ABX task [12, 13], which does not require any training, and only requires a notion of distance between the representations of speech segments.”