Weakly Supervised Multi-Embeddings Learning of Acoustic Models

We trained a Siamese network with multi-task same/different information on a speech dataset, and found that it was possible to share a network for both tasks without a loss in performance. The first task was to discriminate between two same or different words, and the second was to discriminate between two same or different talkers.

cs.SD cs.CL cs.LG Computer Science Linguistics

Paper

References (16)

01Phonetics embedding learning with side informationGabriel Synnaeve, Thomas Schatz, Emmanuel Dupoux2014 · Spoken Language Technology Workshop · 59 citations

02Evaluating speech features with the minimal-pair ABX task: analysis of the classical MFC/PLP pipelineThomas Schatz, Vijayaditya Peddinti, F. Bach et al.2013 · Interspeech · 170 citations

03ADADELTA: An Adaptive Learning Rate MethodMatthew D. Zeiler2012 · arXiv.org · 6.9k citations In Library

04Understanding how Deep Belief Networks perform acoustic modellingAbdel-rahman Mohamed, Geoffrey E. Hinton, Gerald Penn2012 · IEEE International Conference on Acoustics, Speech, and Signal Processing · 322 citations

05At 6–9 months, human infants know the meanings of many common nounsElika Bergelson, Daniel Swingley2012 · Proceedings of the National Academy of Sciences of the United States of America · 849 citations

06Front-End Factor Analysis for Speaker VerificationN. Dehak, P. Kenny, Réda Dehak et al.2011 · IEEE Transactions on Audio, Speech, and Language Processing · 3.6k citations

07Infant ability to tell voices apart rests on language experience.Elizabeth K. Johnson, Ellen Westrek, T. Nazzi et al.2011 · Developmental Science · 119 citations

08Speaker Diarization: A Review of Recent ResearchXavier Anguera Miró, Simon Bozonnet, N. Evans et al.2010 · IEEE Transactions on Audio, Speech, and Language Processing · 758 citations

09Unsupervised learning of vowel categories from infant-directed speechGautam K. Vallabha, James L. McClelland, Ferran Pons et al.2007 · Proceedings of the National Academy of Sciences of the United States of America · 265 citations

10Signature Verification Using A "Siamese" Time Delay Neural NetworkJ. Bromley, J. Bentz, L. Bottou et al.1993 · International journal of pattern recognition and artificial intelligence · 4.2k citations

11Under review as a workshop contribution at ICLR2015 · Under review as a workshop contribution at ICLR

12Dropout: a simple way to prevent neural networks from overfittingNitish Srivastava, Geoffrey E. Hinton, A. Krizhevsky et al.2014 · Journal of machine learning research · 44k citations

Scroll for more · 4 remaining

Similar papers

Learning weakly supervised multimodal phoneme embeddingsR. Chaabouni, Ewan Dunbar, Neil Zeghidour et al.2017 · Interspeech · 12 citations In Library

Additional Shared Decoder on Siamese Multi-view Encoders for Learning Acoustic Word EmbeddingsMyunghun Jung, Hyungjun Lim, Jahyun Goo et al.2019 · Automatic Speech Recognition & Understanding · 16 citations In Library

Discriminative Acoustic Word Embeddings: Recurrent Neural Network-Based ApproachesShane Settle, Karen Livescu2016 · Spoken Language Technology Workshop · 92 citations In Library

Deep convolutional acoustic word embeddings using word-pair side informationH. Kamper, Weiran Wang, Karen Livescu2015 · IEEE International Conference on Acoustics, Speech, and Signal Processing · 181 citations In Library

Sampling strategies in Siamese Networks for unsupervised speech representation learningRachid Riad, Corentin Dancette, Julien Karadayi et al.2018 · Interspeech · 28 citations In Library

Unified Voice Embedding Through Multi-task LearningJenarthanan Rajenthiran, Lakshikka Sithamaparanathan, Sarany Uthayakumar et al.2022 · International Conference on Asian Language Processing · 0 citations In Library

Towards multi-task learning of speech and speaker recognitionNik Vaessen, David A. van Leeuwen2023 · arXiv (Cornell University) · 0 citations In Library

Multi-view Recurrent Neural Acoustic Word EmbeddingsWanjia He, Weiran Wang, Karen Livescu2016 · International Conference on Learning Representations · 93 citations In Library

Comparing supervised and self-supervised embedding for ExVo Multi-Task learning trackTilak Purohit, Imen Ben Mahmoud, Bogdan Vlasenko et al.2022 · arXiv (Cornell University) · 8 citations In Library

Learning Joint Acoustic-Phonetic Word EmbeddingsMohamed El-Geish2019 · arXiv.org · 3 citations In Library

Source of record

arXiv

44B

Welcome.

This panel sticks with you. Pick anything below and it opens right beside it, so you can dig through 60,000-plus records without ever losing your spot here.

Travel 44B

LibraryPapers, policy, standards, statute LabsEvery organization building AI ModelsIntelligence, price, and speed BenchmarksThe evaluation catalog IncidentsAI failures and harms SearchOne field across everything DashboardThe state of AI accountability in NY 44B RegistryThe Art. 44-B compliance portal