Self-Supervised Phonotactic Representations for Language Identification

Ramesh, Gowtham; Kumar, C. Shiva; Murty, K. Sri Rama

doi:10.21437/interspeech.2021-1310

Cited by 8 publications

(5 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Ramesh et al (2021) [147] explored self-supervised approaches to build phonotactic based LID systems for seven Indian languages; Bengali, Hindi, Malayalam, Marathi, Punjabi, Tamil, and Telugu. The audios were collected from video streaming websites.…”

Section: Das Et Al (2020)mentioning

confidence: 99%

“…For the languages with near to no standard resource available, zero-shot learning [196] can be an efective solution. Self-supervised learning [8,147,149,174] is also a promising approach to develop speech technologies with severely low-resourced Indian languages where veriied ground truth labels are very challenging to collect.…”

Section: Issue Of Low-resourcementioning

confidence: 99%

“…However, there are some recent multilingual ASR challenges (discussed in Section 6.3), which attempted to solve the issues of transcripts for the Indian languages. Alternatively, selfsupervised approaches [8,147,174] and transfer learning can also be useful to build large-scale phonotactic Indian LID systems with very limited transcripts.…”

Section: Extracting Language-specific Featuresmentioning

confidence: 99%

See 2 more Smart Citations

An Overview of Indian Spoken Language Recognition from Machine Learning Perspective

Dey

Sahidullah

Saha

2022

ACM Trans. Asian Low-Resour. Lang. Inf. Process.

View full text Add to dashboard Cite

Automatic spoken language identification (LID) is a very important research field in the era of multilingual voice-command-based human-computer interaction (HCI). A front-end LID module helps to improve the performance of many speech-based applications in the multilingual scenario. India is a populous country with diverse cultures and languages. The majority of the Indian population needs to use their respective native languages for verbal interaction with machines. Therefore, the development of efficient Indian spoken language recognition systems is useful for adapting smart technologies in every section of Indian society. The field of Indian LID has started gaining momentum in the last two decades, mainly due to the development of several standard multilingual speech corpora for the Indian languages. Even though significant research progress has already been made in this field, to the best of our knowledge, there are not many attempts to analytically review them collectively. In this work, we have conducted one of the very first attempts to present a comprehensive review of the Indian spoken language recognition research field. In-depth analysis has been presented to emphasize the unique challenges of low-resource and mutual influences for developing LID systems in the Indian contexts. Several essential aspects of the Indian LID research, such as the detailed description of the available speech corpora, the major research contributions, including the earlier attempts based on statistical modeling to the recent approaches based on different neural network architectures, and the future research trends are discussed. This review work will help assess the state of the present Indian LID research by any active researcher or any research enthusiasts from related fields.

show abstract

Section: Das Et Al (2020)mentioning

confidence: 99%

Section: Issue Of Low-resourcementioning

confidence: 99%

Section: Extracting Language-specific Featuresmentioning

confidence: 99%

See 1 more Smart Citation

An Overview of Indian Spoken Language Recognition from Machine Learning Perspective

Dey

Sahidullah

Saha

2022

ACM Trans. Asian Low-Resour. Lang. Inf. Process.

View full text Add to dashboard Cite

show abstract

“…Here, we use the wav2vec2-base-960h pretrained model from 2 without further self-supervised training or supervised fine-tuning, extracting embeddings as the low-level descriptors (LLDs) to form the basis for more elaborate feature representations. In [26], the authors showed that aggregating wav2vec 2.0 embeddings outperforms supervised counterparts, and they show aggregation is suitable for extracting phonotactic constraints. In [21], the authors showed the effectiveness of using different layers from the pretrained wav2vec model on emotion recognition tasks.…”

Section: Transformer-based Acoustic Featuresmentioning

confidence: 99%

Multi-level Fusion of Fisher Vector Encoded BERT and Wav2vec 2.0 Embeddings for Native Language Identification

Krebbers

Kaya

Karpov

2022

Speech and Computer

View full text Add to dashboard Cite

Native Language Identification is a prominent paralinguistic study with applications ranging from biometric analysis to speaker adaptation. Former studies on this task have benefited from alternative acoustic feature representations and pre-trained neural networks. In this work, we explore the Native Language Identification performance of contextual acoustic (wav2vec 2.0) and linguistic (BERT) embeddings as state-of-the-art feature representations and combine them with acoustic features at different levels. We encode acoustic and linguistic features using Fisher Vectors, applying Fisher Vector encoding on BERT word embeddings and wav2vec 2.0 for the first time for a paralinguistic task. We compare this approach with conventional functional summarization. In line with our former study using only acoustic modality, the results indicate the superiority of Fisher Vectors encoding over the traditional techniques. Moreover, we show the efficacy of combining alternative representations now in both acoustic and linguistic modalities. Results indicate a notable contribution of the transformer-based contextual auditory and linguistic feature representations to bimodal Native Language Identification systems.

show abstract

“…The fine-tuned model can be then used as a feature extractor with the extracted representations subsequently being used as input to the downstream model [24,91].…”

Section: Introductionmentioning

confidence: 99%

Enhancing spoken language identification and diarization for multilingual speech

Liu¹

View full text Add to dashboard Cite

Suzy J. Styles for their insightful and generous comments and suggestions on my research. I wish to extend my thanks to Dr. Justin Dauwels for o↵ering me the scholarship that allows me to pursue this degree. I would also like to thank my team members of the NRF and ISSAC projects for a cherished time spent together. I enjoy our stimulating discussion and collaboration in our meetings. Many thanks to all people at Nanyang Technological University who help me to complete my Ph.D. program. I want to thank my friends, whose warm concern and lovely chat have accompanied me these years. Finally, I would like to express my special thanks to my family, especially my mom, for their consistent love and support. I am aware of how dull and di cult I can be. I just want to say, in my way, I love you all.

show abstract

Self-Supervised Phonotactic Representations for Language Identification

Cited by 8 publications

References 0 publications

An Overview of Indian Spoken Language Recognition from Machine Learning Perspective

An Overview of Indian Spoken Language Recognition from Machine Learning Perspective

Multi-level Fusion of Fisher Vector Encoded BERT and Wav2vec 2.0 Embeddings for Native Language Identification

Enhancing spoken language identification and diarization for multilingual speech

Contact Info

Product

Resources

About