Efficient spoken term discovery using randomized algorithms

By Aren Jansen, Benjamin Van Durme. 2011

174 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

Unsupervised Speech Representation Learning Using WaveNet Autoencoders (2019, methodology): “[61] trains an autoencoder on pairs of speech segments found using an unsupervised term discovery system [79].”
Jointly Discovering Visual Objects and Spoken Words from Raw Sensory Input (2018, background): “Popular approaches are based on dynamic time warping [21,22,38], or Bayesian generative models of the speech signal [25,30,35].”
The Zero Resource Speech Challenge 2015: Proposed Approaches and Results (2016, methodology): “As the baseline, we evaluated a previously existing spoken term discovery system [18].”
Unsupervised Cross-Modal Alignment of Speech and Text Embedding Spaces (2018, background): “Early work mainly focused on un-supervised 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): “Term discovery systems like JHU-PLP [15] aim to find high-precision clusters of isolated segments, but do not cover all the data.”
Learning Word-Like Units from Joint Audio-Visual Analysis (2017, methodology): “Previous work has presented algorithms for performing acoustic pattern discovery in continuous speech (Park & Glass, 2008; Jansen et al., 2010; Jansen & Van Durme, 2011) without the use of transcriptions or another modality, but those algorithms are limited in their ability to scale by their inheren”
The zero resource speech challenge 2015 (2015, methodology): “As baseline, we used the word discovery system described in [8] run on PLP features.”
Linguistic Unit Discovery from Multi-Modal Inputs in Unwritten Languages: Summary of the “Speaking Rosetta” JSALT 2017 Workshop (2018, methodology): “System Prec Recall F Segmental DTW Baseline [37] 31.”
An empirical evaluation of zero resource acoustic unit discovery (2017, background): “Since such word pairs can not only be obtained from manual transcripts, also from unsupervised spoken term discovery systems without relying on any language-specific resources [12, 13], at little cost compared with expensive phoneme transcripts.”
Virtual Phone Discovery for Speech Synthesis Without Text (2019, methodology): “System Boundary Precision Recall F-score Baseline [31] 44.”
The Zero Resource Speech Challenge 2020: Discovering discrete subword and word units (2020, methodology): “The baseline was computed using [21], which does pair-matching using locally sensitive hashing applied to PLP features and then groups pairs using graph clustering.”
Unsupervised neural network based feature extraction using weak top-down constraints (2015, methodology): “In our truly unsupervised setup, we use word pairs discovered using the UTD system of [12].”
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”
A summary of the 2012 JHU CLSP workshop on zero resource speech technologies and models of early language acquisition (2013, unknown): “The word discovery system, described in [38], produces a list of speech interval pairs (also with a confidence score), that likely contain the same word or phrase, but whose identity is unknown.”
Spoken Content Retrieval—Beyond Cascading Speech Recognition with Text Retrieval (2015, methodology): “A very efficient frame indexing approach was proposed for this purpose [229], [230] by applying locality sensitive hashing techniques on the frames [231], which was shown to be a good approximation for the cosine similarity.”
Unsupervised Word Segmentation and Lexicon Discovery Using Acoustic Word Embeddings (2016, methodology): “Examples include the query-by-example systems of [1]–[4], and the unsupervised term discovery (UTD) systems of [5], [6], which aim to ﬁnd repeated words or phrases in a speech collection.”
Weak top-down constraints for unsupervised acoustic model training (2013, methodology): “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].”