Use of articulatory bottle-neck features for query-by-example spoken term detection in low resource scenarios

Mantena, Gautam Varma; Prahallad, Kishore

doi:10.1109/icassp.2014.6854983

Cited by 17 publications

(5 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Image processing techniques are also widely employed for verifying the presence of the query in an utterance [37–39 ]. Research studies in STD with focus on low resourced languages are in progress [16, 19 ].…”

Section: Related Work On Stdmentioning

confidence: 99%

“…However, the development of LVCSR system is possible only for well‐resourced languages, and it has difficulty in handling out‐of‐vocabulary (OOV) words. Later, predictive neural networks [12 ], phone lattice alignments [13, 14 ], multilayer perceptrons [15, 16 ], and deep neural networks (DNNs) [17–21 ] were introduced to keyword spotting. Variety features such as spectrographic seam patterns [22 ] and spectro–temporal patch features [23 ] have also experimented.…”

Section: Related Work On Stdmentioning

confidence: 99%

“…Most of the unsupervised techniques employ variants of DTW for verifying the query occurrences at the later template matching stage. Besides standard DTW algorithm, its variants like unconstrained‐endpoint DTW [34 ], modified DTW [28 ], segmental DTW (S‐DTW) [30, 31 ], modified S‐DTW [35, 36 ], and non‐segmental DTW (NS‐DTW) [16, 32, 37 ] is also used. Image processing techniques are also widely employed for verifying the presence of the query in an utterance [37–39 ].…”

Section: Related Work On Stdmentioning

confidence: 99%

See 2 more Smart Citations

Two‐stage spoken term detection system for under‐resourced languages

Deekshitha

Mary²

2020

IET signal process.

View full text Add to dashboard Cite

Section: Related Work On Stdmentioning

confidence: 99%

Section: Related Work On Stdmentioning

confidence: 99%

Section: Related Work On Stdmentioning

confidence: 99%

See 1 more Smart Citation

Two‐stage spoken term detection system for under‐resourced languages

Deekshitha

Mary²

2020

IET signal process.

View full text Add to dashboard Cite

“…To deal with OOV problem, many approaches using a feature-based acoustic match have shown its effectiveness in low-resource STD tasks, as well as the robustness against the effects by difference of speaker and recording environments [1], [2], [3]. However, a feature-based approach is time-consuming and the approach alone couldn't outperform the STD performance of conventional ASR-based system for rich-resource language tasks.…”

Section: Introductionmentioning

confidence: 99%

Combining State-Level Spotting and Posterior-Based Acoustic Match for Improved Query-by-Example Spoken Term Detection

Oishi¹,

Matsuba²,

Makino³

et al. 2016

Interspeech 2016

View full text Add to dashboard Cite

In spoken term detection (STD) systems, automatic speech recognition (ASR) frontend is often employed for its reasonable accuracy and efficiency. However, out-of-vocabulary (OOV) problem at ASR stage has a great impact on the STD performance for spoken query. In this paper, we propose combining feature-based acoustic match which is often employed in the STD systems for low resource languages, along with the other ASR-derived features. First, automatic transcripts for spoken document and spoken query are decomposed into corresponding acoustic model state sequences and used for spotting plausible speech segments. Second, DTW-based acoustic match between the query and candidate segment is performed using the posterior features derived from a monophone-state DNN. Finally, an integrated score is obtained by a logistic regression model, which is trained with a large spoken document and automatically generated spoken queries as development data. The experimental results on NTCIR-12 SpokenQuery&Doc-2 task showed that the proposed method significantly outperforms the baseline systems which use the subword-level or state-level spotting alone. Also, our universal scoring model trained with a separate set of development data could achieve the best STD performance, and showed the effectiveness of additional ASRderived features regarding the confidence measure and query length.

show abstract

“…The most widely used methods rely on template matching of speech features (eg: Gaussian or phone posteriorgrams) using dynamic time warping (DTW) algorithms [6]- [8], and similarity search based on end-to-end neural networks [9]. There exist hybrid approaches that use discriminatively trained models to extract speech features (eg: multilingual, articulatory bottleneck features) [10]- [13], which are then used in a template matching framework.…”

mentioning

confidence: 99%

Resources and Benchmarks for Keyword Search in Spoken Audio From Low-Resource Indian Languages

et al. 2022

View full text Add to dashboard Cite

This paper presents the resources and benchmarks developed for keyword search (KWS) in spoken audio from six low-resource Indian languages (from two families), namely Gujarati, Hindi, Marathi, Odia, Tamil, and Telugu. The current work on constructing keywords and building benchmark KWS systems is inspired by the popular IARPA Babel program and the subsequent works on low-resource KWS. The keywords are constructed by taking into account their properties i.e., occurrence, length, and average confusability; and their effects on the evaluation metric -the term-weighted value (TWV). We make use of freely available speech datasets, and reprocess them to create resources for KWS, thereby adding value to the existing speech resources. Four ASR-based KWS systems are built, and their performance is analyzed across the three keyword properties on all the six languages. The prepared keywords and other related resources to replicate our experiments are made available for the public. We believe that the analysis and guidelines provided in this paper will not only help the research community, but also practitioners and engineers to easily create KWS resources for newer languages, datasets, and scenarios.INDEX TERMS Keyword search, low-resource languages, term-weighted value (TWV).

show abstract

Use of articulatory bottle-neck features for query-by-example spoken term detection in low resource scenarios

Cited by 17 publications

References 22 publications

Two‐stage spoken term detection system for under‐resourced languages

Two‐stage spoken term detection system for under‐resourced languages

Combining State-Level Spotting and Posterior-Based Acoustic Match for Improved Query-by-Example Spoken Term Detection

Resources and Benchmarks for Keyword Search in Spoken Audio From Low-Resource Indian Languages

Contact Info

Product

Resources

About