Unsupervised Pattern Discovery in Speech

By Alex Seungryong Park, James R. Glass. 2008

365 citations recorded by the source corpus.

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

In Signal Processing

Citations in the corpus

Learning Hierarchical Discrete Linguistic Units from Visually-Grounded Speech (2019, methodology): “One family of techniques is based upon Segmental Dynamic Time Warping (S-DTW) (Park & Glass, 2005; 2008; Jansen et al., 2010; Jansen & Van Durme, 2011), which uses a self-comparison algorithm to identify relatively long duration (on the order of a second) patterns which frequently reoccur in a speec”
Effectiveness of self-supervised pre-training for speech recognition (2019, background): “The practical need for building ASR systems for new conditions with limited resources spurred a lot of work focused on unsupervised speech recognition and representation learning (Park and Glass, 2008; Jansen et al.; Glass; et. al., a,f; Chung and Glass; van den Oord et al., 2018; Schneider et al.)”
Unsupervised Speech Representation Learning Using WaveNet Autoencoders (2019, background): “Linguistic unit discovery models detect recurring speech patterns at a word-like level, finding commonly repeated segments with a constrained dynamic time warping [74].”
A segmental framework for fully-unsupervised large-vocabulary speech recognition (2016, methodology): “…in tasks such as query-by-example search [16, 17], where a system needs to ﬁnd all the utterances in a corpus containing a spoken query, or in unsupervised term discovery (UTD), where a system needs to automatically ﬁnd repeated word-or phrase-like patterns in a speech collection [1, 18, 19].”
Unsupervised spoken keyword spotting via segmental DTW on Gaussian posteriorgrams (2009, methodology): “Given spoken samples of a keyword, we apply the segmental dynamic time warping (SDTW) that we have explored previously [ 18 ] to compare the Gaussian posteriorgrams between keyword samples and the test utterances.”
Unsupervised Cross-Modal Alignment of Speech and Text Embedding Spaces (2018, background): “Early work mainly focused on unsupervised term discovery, where the aim is to find word- or phrase-like patterns in a collection of speech [29, 30].”
An embedded segmental K-means model for unsupervised segmentation and clustering of speech (2017, background): “Early work mainly focused on unsupervised term discovery, where the aim is to automatically ﬁnd repeated word-or phrase-like patterns in a collection of speech [14–16].”
Learning Word-Like Units from Joint Audio-Visual Analysis (2017, methodology): “Segmental Dynamic Time Warping (S-DTW) was introduced by Park & Glass (2008), which discovers repetitions of the same words and phrases in a collection of untran-scribed acoustic data.”
Unsupervised Speech Recognition via Segmental Empirical Output Distribution Matching (2018, background): “The standard approaches segment audio signals that are acoustically similar, and cluster the obtained segmented signals (Lee and Glass, 2012; Glass, 2012; Park and Glass, 2008; Driesen et al., 2012).”
Effectiveness of Self-Supervised Pre-Training for ASR (2020, background): “The practical need for building ASR systems for new conditions with limited resources spurred a lot of work focused on unsupervised speech recognition and representation learning [2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14], in addition to semi- and weakly-supervised learning techniques to reduce th”
Spoken Content Retrieval—Beyond Cascading Speech Recognition with Text Retrieval (2015, methodology): “1) Bottom-up modeling: Most approaches for modeling the acoustic patterns follow a three-phase recursive procedure including signal segmentation, unit clustering and model training in each iteration [85], [202], [209], [216], [233]–[239].”
The zero resource speech challenge 2015 (2015, result): “These two levels have already been investigated in previous work (see [1, 2, 3, 4, 5, 6, 7], and [8, 9, 10, 11], respectively), but the performance of the different systems has not yet been compared using common evaluation metrics and data sets.”
Fully unsupervised small-vocabulary speech recognition using a segmental Bayesian model (2015, methodology): “Walter et al. [18] used both the boundaries and word identities for initialization in the the UTD case, while we use only the word boundaries for initialization, and leave it up to the model to discover the clusters.”
Linguistic Unit Discovery from Multi-Modal Inputs in Unwritten Languages: Summary of the “Speaking Rosetta” JSALT 2017 Workshop (2018, background): “One strand of research focuses on discovering the linguistic units of the low-resource language from the raw speech data, while assuming no other information about the language is available, and using these to build ASR systems (zero resource approach; e.g., [2, 3, 4, 5, 6]).”
Weak top-down constraints for unsupervised acoustic model training (2013, background): “We find these weak top-down constraints are capable of improving model speaker independence by up to 57% relative over bottom-up training alone.”
Unsupervised Learning of Spoken Language with Visual Context (2016, methodology): “The embeddings might be combined with other forms of word clustering such as those based on dynamic time warping [3] to perform acoustically and semantically aware word clustering.”