Towards spoken term discovery at scale with zero resources

By Aren Jansen, Kenneth Ward Church, Hynek Hermansky. 2010

142 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…”
Effectiveness of self-supervised pre-training for speech recognition (2019, methodology): “Our work is inspired by research efforts reducing the dependence on labeled data for building ASR systems through unsupervised unit discovery and acoustic representation leaning [5, 6, 7, 8, 9], and through multi- and cross-lingual transfer learning in low-resource conditions [46, 47, 48, 49, 50, 51”
Jointly Discovering Visual Objects and Spoken Words from Raw Sensory Input (2018, background): “Popular approaches are based on dynamic time warping (Jansen et al. 2010; Jansen and Van Durme 2011; Park and Glass 2008), or Bayesian generative models of the speech signal (Kamper et al. 2016; Lee and Glass 2012; Ondel et al. 2016).”
Fixed-dimensional acoustic embeddings of variable-length segments in low-resource settings (2013, background): “Such embeddings would (i) enable the application of standard distance learning techniques [5, 6] to template-based acoustic modeling and (ii) support a new generation of efficient segment-based audio indexing algorithms, enabling highly scalable spoken term discovery [7, 8, 9] and query-by-example s”
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”
A summary of the 2012 JHU CLSP workshop on zero resource speech technologies and models of early language acquisition (2013, methodology): “Speech recognition researchers have developed discovery systems that search for repeated acoustic patterns, treating the remainder of the corpus as background [2, 3].”
Unsupervised neural network based feature extraction using weak top-down constraints (2015, background): “Zero-resource technology aims to solve tasks such as phonetic and lexical discovery [4, 5], spoken document retrieval [6], and query-by-example search [7, 8] by using only raw speech data.”
Deep learning methods for unsupervised acoustic modeling — Leap submission to ZeroSpeech challenge 2017 (2017, background): “In the recent years, there has been a growing interest in the task of unsupervised representation learning from raw speech data without any supervision or labels [1, 2, 3].”
Towards Visually Grounded Sub-word Speech Unit Discovery (2019, background): “Some proposed approaches operate at the word level [1, 2, 3], while a separate line of work is concerned with sub-word modeling [4, 5, 6, 7].”
Pre-training on high-resource speech recognition improves low-resource speech-to-text translation (2018, result): “This hypothesis is supported by other work showing the benefits of cross-lingual and multilingual training for speech technology in low-resource target languages (Carlin et al., 2011; Jansen et al., 2010; Deng et al., 2013; Vu et al., 2012; Thomas et al., 2012; Cui et al., 2015; Alumäe et al., 2016;”
Segmental acoustic indexing for zero resource keyword search (2015, background): “Dynamic time warping (DTW) has been effective in zeroresource query-by-example search [8, 9, 10].”
Weak top-down constraints for unsupervised acoustic model training (2013, background): “However, the past five years have seen several efforts in developing scalable automatic approaches to discovering repeated words and phrases in large speech corpora using nothing but the raw feature vectors as input [1, 2, 3, 4, 5, 6].”
Spoken Content Retrieval—Beyond Cascading Speech Recognition with Text Retrieval (2015, methodology): “In this approach, segmental DTW mentioned in Subsection V-B was employed first to discover a set of audio intervals, and similar audio intervals very probably sharing the same underlying text transcription were linked together [234].”
Efficient spoken term discovery using randomized algorithms (2011, methodology): “useful for summarization [6][8] and speech retrieval applications [9][10].”
Unsupervised training of an HMM-based self-organizing unit recognizer with applications to topic classification and keyword discovery (2014, methodology): “An alternative approach for the unsupervised modeling of speech was originally developed by Park and Glass 2005) and more recently by Jansen et al. (2010) which is a non-parametric approach that is related to the Dynamic ime Warping (DTW) technology for speech recognition.”