Modern Speech Recognition for Romanian Language
Source abstract
Abstract
Despite having approximately 24 million native speakers, Romanian remains a low-resource language for automatic speech recognition (ASR), with few accurate and publicly available systems. To address this gap, this study explores the challenges of adapting modern speech recognition models, such as wav2vec 2.0 and Conformer, to Romanian. Our investigation is a comprehensive analysis of the two models, their capabilities to adapt to Romanian data, and the performance of the trained models. The research also focuses on unique attributes of the Romanian language, data collection techniques, including weakly supervised learning, and processing methodologies. Building on the previously introduced Echo dataset of 378 h, we release CRoWL (Crawled Romanian Weakly Labeled), a weakly supervised dataset of 9000 h created via automatic transcription. We obtain strong results that, to the best of our knowledge, are competitive with or exceed publicly reported results for Romanian under comparable open evaluation settings, with Conformer attaining 3.01% WER on Echo + CRoWL and wav2vec 2.0 reaching 4.04% (Echo) and 4.17% (Echo + CRoWL). In addition to the datasets, we also release our most capable models as open source, along with their training plans, thereby providing a solid foundation for researchers interested in languages with limited representation.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Speech Recognition and Synthesis · Speech and Audio Processing · Phonetics and Phonology Research
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
Computer Science & Engineering Department, National University of Science and Technology POLITEHNICA Bucharest, 060042 Bucharest, Romania
Science and Information Technology Section, Academy of Romanian Scientists, Ilfov 3, 050044 Bucharest, Romania
Funding
Smart Growth, Digitization and Financial Instruments Program
References (24)
Show source-provided references
1. Eberhard, D.M., Simons, G.F., and Fennig, C.D. (2023). Ethnologue: Languages of the World, SIL International (SIL Global Publishing). [26th ed.].
2. Posner, R. (2026, February 07). Romanian Language. Encyclopaedia Britannica, Available online: https://www.britannica.com/topic/Romanian-language.
3. Ungureanu, D., Toma, S.A., Filip, I.D., Mocanu, B.C., Aciobăniței, I., Marghescu, B., Balan, T., Dascalu, M., Bica, I., and Pop, F. (2023). ODIN112–AI-Assisted Emergency Services in Romania. Appl. Sci., 13.10.3390/app13010639 ↗
4. Ungureanu, D., Ruseti, S., Toma, I., and Dascalu, M. (2022). pROnounce: Automatic Pronunciation Assessment for Romanian. Conference on Smart Learning Ecosystems and Regional Development, Springer.10.1007/978-981-19-5240-1_7 ↗
5. Echo: A Crowd-sourced Romanian Speech Dataset
6. Chitoran, I. (2013). The Phonology of Romanian: A Constraint-Based Approach, Walter de Gruyter.
7. Pană Dindelegan, G. (2013). The Grammar of Romanian, Oxford University Press.
8. Renwick, M.E. (2012). Vowels of Romanian: Historical, Phonological and Phonetic Studies. [Ph.D. Thesis, Cornell University].
9. Pană Dindelegan, G. (2013). Inflectional and Derivational Morphophonological Alternations. The Grammar of Romanian, Oxford University Press.
10. Semi-supervised Approach to Romanian Noun Declension10.1016/j.procs.2016.08.248 ↗
11. Paskaleva, E., and Slavcheva, M. (2007). Diacritics Restoration in Romanian Texts. Proceedings of the a Common Natural Language Processing Paradigm for Balkan Languages—RANLP 2007 Workshop Proceedings, Borovets, Bulgaria, 27–29 September 2007, INCOMA Ltd.
12. Roseano, P., Turculeţ, A., Bibiri, A.D., Cerdà Massó, R., Fernández Planas, A.M., and Elvira-García, W. (2022). A dialectometric approach to Romanian intonation. Onomázein, 105–139.10.7764/onomazein.ne11.07 ↗
13. wav2vec 2.0: A framework for self-supervised learning of speech representations
14. Gulati, A., Qin, J., Chiu, C.C., Parmar, N., Zhang, Y., Yu, J., Han, W., Wang, S., Zhang, Z., and Wu, Y. (2020). Conformer: Convolution-augmented transformer for speech recognition. arXiv.10.21437/Interspeech.2020-3015 ↗
15. Babu, A., Wang, C., Tjandra, A., Lakhotia, K., Xu, Q., Goyal, N., Singh, K., Von Platen, P., Saraf, Y., and Pino, J. (2021). XLS-R: Self-supervised cross-lingual speech representation learning at scale. arXiv.10.21437/Interspeech.2022-143 ↗
16. Rekesh, D., Koluguri, N.R., Kriman, S., Majumdar, S., Noroozi, V., Huang, H., Hrinchuk, O., Puvvada, K., Kumar, A., and Balam, J. (2023). Fast conformer with linearly scalable attention for efficient speech recognition. Proceedings of the 2023 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU), IEEE.10.1109/ASRU57964.2023.10389701 ↗
17. Amodei, D., Ananthanarayanan, S., Anubhai, R., Bai, J., Battenberg, E., Case, C., Casper, J., Catanzaro, B., Cheng, Q., and Chen, G. (2016). Deep speech 2: End-to-end speech recognition in english and mandarin. Proceedings of the International Conference on Machine Learning, International Machine Learning Society (IMLS).
18. Radford, A., Kim, J.W., Xu, T., Brockman, G., McLeavey, C., and Sutskever, I. (2023, January 23–29). Robust speech recognition via large-scale weak supervision. Proceedings of the International Conference on Machine Learning (PMLR), Honolulu, HI, USA.
19. An automatic speech recognition system with speaker-independent identification support
20. Georgescu, A.L., Cucu, H., Buzo, A., and Burileanu, C. (2020, January 11–16). RSC: A Romanian read speech corpus for automatic speech recognition. Proceedings of the Twelfth Language Resources and Evaluation Conference, Marseille, France.
21. Stan, A., Dinescu, F., Ţiple, C., Meza, Ş., Orza, B., Chirilă, M., and Giurgiu, M. (2017). The SWARA speech corpus: A large parallel Romanian read speech dataset. Proceedings of the 2017 International Conference on Speech Technology and Human-Computer Dialogue (SpeD), IEEE.10.1109/SPED.2017.7990428 ↗
22. Ardila, R., Branson, M., Davis, K., Henretty, M., Kohler, M., Meyer, J., Morais, R., Saunders, L., Tyers, F.M., and Weber, G. (2019). Common voice: A massively-multilingual speech corpus. arXiv.
23. Georgescu, A.L., Cucu, H., and Burileanu, C. (2019). Kaldi-based DNN architectures for speech recognition in Romanian. Proceedings of the 2019 International Conference on Speech Technology and Human-Computer Dialogue (SpeD), IEEE.10.1109/SPED.2019.8906555 ↗
24. Park, D.S., Chan, W., Zhang, Y., Chiu, C.C., Zoph, B., Cubuk, E.D., and Le, Q.V. (2019). SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition. Interspeech 2019, ISCA.10.21437/Interspeech.2019-2680 ↗