The NEUREM3 project encompasses basic research in speech processing (SP) and natural language processing (NLP) with accent on multi-linguality and multi-modality (speech and text processing with the support of visual information). Current deep machine learning methods are based on continuous vector representations that are created by the neural networks (NN) themselves during the training. Although empirically, the results of such NNs are often excellent, our knowledge and understanding of such representations is insufficient. NEUREM3 has an ambition to fill this gap and to study neural representations for speech and text units of different scopes (from phonemes and letters to whole spoken and written documents) and representations acquired both for isolated tasks and multi-task setups. NEUREM3 will also improve NN architectures and training techniques, so that they can be trained on incomplete or incoherent data.

Description in Czech
Projekt NEUREM3 spojuje základní výzkum v oblasti zpracování mluvené řeči (speech processing, SP) a přirozeného jazyka (natural language processing, NLP) s důrazem na vícejazyčnost a multi-modalitu (zpracování řeči a textu s podporou obrazové informace). V jádru současných metod hlubokého strojového učení leží spojité vektorové reprezentace, které si neuronové samy budují během trénování. Ačkoli empiricky dosahují neuronové sítě často vynikajících výsledků, znalosti a pochopení získaných reprezentací jsou nedostatečné. NEUREM3 má ambici tuto mezeru vyplnit a studovat neuronové reprezentace pro jednotky textu a řeči různého rozsahu (od fonémů a písmen až po proslovy a dokumenty) a reprezentace získané pro izolované úlohy i více úloh současně (multi-tasking). NEUREM3 vylepší architektury i techniky trénování neuronových sítí, aby je bylo možné trénovat je na neúplných nebo nekoherentních datech.

Keywords
deep learning;machine learning;neural networks;continuous representations;natural language processing;speech and text processing;machine translation;multi-modality;multi-linguality

Key words in Czech
hluboké strojové učení;neuronové sítě;spojité reprezentace;zpracování přirozeného jazyka;zpracování řeči a textu;strojový překlad; multimodalita;mnohojazyčnost

GX19-26934X

English

Burget Lukáš, doc. Ing., Ph.D. - principal person responsible
Baskar Murali Karthick, Ing., Ph.D. - fellow researcher
Beneš Karel, Ing. - fellow researcher
Han Jiangyu - fellow researcher
Kesiraju Santosh, Ph.D. - fellow researcher
Peng Junyi - fellow researcher
Plchot Oldřich, Ing., Ph.D. - fellow researcher
Rohdin Johan Andréas, M.Sc., Ph.D. - fellow researcher
Sarvaš Marek, Ing. - fellow researcher

Department of Computer Graphics and Multimedia
- responsible department (8.6.2018 - not assigned)
Speech Data Mining Research Group BUT Speech@FIT
- internal (8.6.2018 - 31.12.2023)
Department of Computer Graphics and Multimedia
- beneficiary (8.6.2018 - 31.12.2023)

LANDINI, F.; DIEZ SÁNCHEZ, M.; STAFYLAKIS, T.; BURGET, L. DiaPer: End-to-End Neural Diarization With Perceiver-Based Attractors. IEEE Transactions on Audio, Speech, and Language Processing, 2024, vol. 32, no. 7, p. 3450-3465. ISSN: 1558-7916.
Detail

KLEMENT, D.; DIEZ SÁNCHEZ, M.; LANDINI, F.; BURGET, L.; SILNOVA, A.; DELCROIX, M.; TAWARA, N. Discriminative Training of VBx Diarization. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Seoul: IEEE Signal Processing Society, 2024. p. 11871-11875. ISBN: 979-8-3503-4485-1.
Detail

PENG, J.; DELCROIX, M.; OCHIAI, T.; ASHIHARA, T.; PLCHOT, O.; ARAKI, S.; ČERNOCKÝ, J. Probing Self-Supervised Learning Models With Target Speech Extraction. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Seoul: IEEE Signal Processing Society, 2024. p. 535-539. ISBN: 979-8-3503-7451-3.
Detail

PENG, J.; DELCROIX, M.; OCHIAI, T.; PLCHOT, O.; ARAKI, S.; ČERNOCKÝ, J. Target Speech Extraction with Pre-Trained Self-Supervised Learning Models. ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Seoul: IEEE Signal Processing Society, 2024. p. 10421-10425. ISBN: 979-8-3503-4485-1.
Detail

HAN, J.; LANDINI, F.; ROHDIN, J.; DIEZ SÁNCHEZ, M.; BURGET, L.; CAO, Y.; LU, H.; ČERNOCKÝ, J. Diacorrect: Error Correction Back-End for Speaker Diarization. In ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Seoul: IEEE Signal Processing Society, 2024. p. 11181-11185. ISBN: 979-8-3503-4485-1.
Detail

BENEŠ, K.; KOCOUR, M.; BURGET, L. Hystoc: Obtaining Word Confidences for Fusion of End-To-End ASR Systems. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Seoul: IEEE Signal Processing Society, 2024. p. 11276-11280. ISBN: 979-8-3503-4485-1.
Detail

PENG, J.; PLCHOT, O.; STAFYLAKIS, T.; MOŠNER, L.; BURGET, L.; ČERNOCKÝ, J. Improving Speaker Verification with Self-Pretrained Transformer Models. In Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH. Proceedings of Interspeech. Dublin: International Speech Communication Association, 2023. p. 5361-5365. ISSN: 1990-9772.
Detail

MATĚJKA, P.; SILNOVA, A.; SLAVÍČEK, J.; MOŠNER, L.; PLCHOT, O.; KLČO, M.; PENG, J.; STAFYLAKIS, T.; BURGET, L. Description and Analysis of ABC Submission to NIST LRE 2022. In Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH. Proceedings of Interspeech. Dublin: International Speech Communication Association, 2023. p. 511-515. ISSN: 1990-9772.
Detail

DELCROIX, M.; TAWARA, N.; DIEZ SÁNCHEZ, M.; LANDINI, F.; SILNOVA, A.; OGAWA, A.; NAKATANI, T.; BURGET, L.; ARAKI, S. Multi-Stream Extension of Variational Bayesian HMM Clustering (MS-VBx) for Combined End-to-End and Vector Clustering-based Diarization. In Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH. Proceedings of Interspeech. Dublin: International Speech Communication Association, 2023. p. 3477-3481. ISSN: 1990-9772.
Detail

KESIRAJU, S.; SARVAŠ, M.; PAVLÍČEK, T.; MACAIRE, C.; CIUBA, A. Strategies for Improving Low Resource Speech to Text Translation Relying on Pre-trained ASR Models. In Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH. Proceedings of Interspeech. Dublin: International Speech Communication Association, 2023. p. 2148-2152. ISSN: 1990-9772.
Detail

MOŠNER, L.; PLCHOT, O.; PENG, J.; BURGET, L.; ČERNOCKÝ, J. Multi-Channel Speech Separation with Cross-Attention and Beamforming. In Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH. Proceedings of Interspeech. Dublin: International Speech Communication Association, 2023. p. 1693-1697. ISSN: 1990-9772.
Detail

YU, D.; GONG, Y.; PICHENY, A.; RAMABHADRAN, B.; HAKKANI-TÜR, D.; PRASAD, R.; ZEN, H.; SKOGLUND, J.; ČERNOCKÝ, J.; BURGET, L.; MOHAMED, A. Twenty-Five Years of Evolution in Speech and Language Processing. IEEE SIGNAL PROCESSING MAGAZINE, 2023, vol. 40, no. 5, p. 27-39. ISSN: 1558-0792.
Detail

YUSUF, B.; ČERNOCKÝ, J.; SARAÇLAR, M. End-to-End Open Vocabulary Keyword Search With Multilingual Neural Representations. IEEE/ACM TRANSACTIONS ON AUDIO, SPEECH AND LANGUAGE PROCESSING, 2023, vol. 31, no. 08, p. 3070-3080. ISSN: 2329-9290.
Detail

KAKOUROS, S.; STAFYLAKIS, T.; MOŠNER, L.; BURGET, L. Speech-Based Emotion Recognition with Self-Supervised Models Using Attentive Channel-Wise Correlations and Label Smoothing. In Proceedings of ICASSP 2023. Rhodes Island: IEEE Signal Processing Society, 2023. p. 1-5. ISBN: 978-1-7281-6327-7.
Detail

PENG, J.; STAFYLAKIS, T.; GU, R.; PLCHOT, O.; MOŠNER, L.; BURGET, L.; ČERNOCKÝ, J. Parameter-Efficient Transfer Learning of Pre-Trained Transformer Models for Speaker Verification Using Adapters. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. Rhodes Island: IEEE Signal Processing Society, 2023. p. 1-5. ISBN: 978-1-7281-6327-7.
Detail

SILNOVA, A.; BRUMMER, J.; SWART, A.; BURGET, L. Toroidal Probabilistic Spherical Discriminant Analysis. In Proceedings of ICASSP 2023. Rhodes Island: IEEE Signal Processing Society, 2023. p. 1-5. ISBN: 978-1-7281-6327-7.
Detail

KESIRAJU, S.; BENEŠ, K.; TIKHONOV, M.; ČERNOCKÝ, J. BUT Systems for IWSLT 2023 Marathi - Hindi Low Resource Speech Translation Task. In 20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference. Toronto (in-person and online): Association for Computational Linguistics, 2023. p. 227-234. ISBN: 978-1-959429-84-5.
Detail

LANDINI, F.; DIEZ SÁNCHEZ, M.; LOZANO DÍEZ, A.; BURGET, L. Multi-Speaker and Wide-Band Simulated Conversations as Training Data for End-to-End Neural Diarization. In Proceedings of ICASSP 2023. Rhodes Island: IEEE Signal Processing Society, 2023. p. 1-5. ISBN: 978-1-7281-6327-7.
Detail

SILNOVA, A.; SLAVÍČEK, J.; MOŠNER, L.; KLČO, M.; PLCHOT, O.; MATĚJKA, P.; PENG, J.; STAFYLAKIS, T.; BURGET, L. ABC System Description for NIST LRE 2022. Proceedings of NIST LRE 2022 Workshop. Washington DC: National Institute of Standards and Technology, 2023. p. 1-5.
Detail

STAFYLAKIS, T.; MOŠNER, L.; KAKOUROS, S.; PLCHOT, O.; BURGET, L.; ČERNOCKÝ, J. Extracting speaker and emotion information from self-supervised speech models via channel-wise correlations. In 2022 IEEE Spoken Language Technology Workshop, SLT 2022 - Proceedings. Doha: IEEE Signal Processing Society, 2023. p. 1136-1143. ISBN: 978-1-6654-7189-3.
Detail

Responsibility: Burget Lukáš, doc. Ing., Ph.D.