Low Resource ASR: The Surprising Effectiveness of High Resource Transliteration

Khare, Shreya; Mittal, Anshul; Diwan, Anuj; Sarawagi, Sunita; Jyothi, Preethi; Bharadwaj, Samarth

doi:10.21437/interspeech.2021-2062

Cited by 21 publications

(10 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Recent speech recognition personalization efforts like Google's Project Euphonia [40,74] are promising. Similarly, efforts that attempt cross-lingual knowledge transfer in speech recognition from high-to low-resource languages (e.g., [58,116]) can make speech input more inclusive. Even then, the challenge of automatically extracting activity labels, timing, and effort levels from verbal reports remains.…”

Section: Limitations and Future Workmentioning

confidence: 99%

MyMove: Facilitating Older Adults to Collect In-Situ Activity Labels on a Smartwatch with Speech

Kim

Chou

Lee

et al. 2022

CHI Conference on Human Factors in Computing Systems

View full text Add to dashboard Cite

show abstract

Section: Limitations and Future Workmentioning

confidence: 99%

MyMove: Facilitating Older Adults to Collect In-Situ Activity Labels on a Smartwatch with Speech

Kim

Chou

Lee

et al. 2022

CHI Conference on Human Factors in Computing Systems

View full text Add to dashboard Cite

show abstract

“…(4) 基于迁移学习的语音识别知识共享技术利用具有丰富资源的多种语言的语音识别声学建模数据, 进行基于迁移学习的多语种声学建模, 是解决小语种数据资源受限的一种途径 [24,25] . 迁移学习可以通过借鉴丰富资源领域的知识来提升资源匮乏领域的模型性能, 是机器学习领域内广为使用的方法.…”

Section: 利用无标注数据和低资源语音建模技术减少声学智能系统对数据的依赖unclassified

Thinking and exploration into intellectualization of speech and medical acoustics

Yan¹,

Cheng²

2022

Sci. Sin.-Phys. Mech. Astron.

View full text Add to dashboard Cite

As the artificial intelligence boom sweep across all walks of life, researchers are also intellectualizing the area of acoustics. Taking speech and medical acoustics as examples, we first analyzed the significant challenges faced by acoustic intelligence, including data scarcity, algorithm independence, computing power dependence, security, and talent shortage. We also summarized important relevant technology and discussed prospective remedies. Furthermore, we reviewed the recent major scientific advances in speech and medical acoustics, both at home and abroad, and summarized the development history of the domestic intelligent voice industries. Finally, we look forward to the directions and trends of intellectual speech and medical acoustics and provide suggestions for specific issues in these fields.

show abstract

“…Especially sharing of sub-word vocabulary among related languages (of the same or similar families) is of more importance to exploit the inter-relationships between the languages (Khemchandani et al, 2021), so that resource sharing from high-resource languages to or any common script, especially in speech-to-text (Javed et al, 2021) and text-to-speech (Zhang et al, 2021) to obtain a universal representation of text across any language/script. In the case of South Asian languages, it is more convenient to map all scripts to a common Indic script (like Devanagari) which is capable of representing all phonemes used in the Indic families (Khare et al, 2021).…”

Section: Related Workmentioning

confidence: 99%

Unified NMT models for the Indian subcontinent, transcending script-barriers

N.c.¹

2022

Proceedings of the Third Workshop on Deep Learning for Low-Resource Natural Language Processing

View full text Add to dashboard Cite

Highly accurate machine translation systems are very important in societies and countries where multilinguality is very common, and where English often does not suffice. The Indian subcontinent (or South Asia) is such a region, with all the Indic languages currently being under-represented in the NLP ecosystem. It is essential to thoroughly explore various techniques to improve the performance of such lowresource languages at least using the data available in open-source, which itself is something not very explored in the Indic ecosystem. In our work, we perform a study with a focus on improving the performance of very-low-resource South Asian languages, especially of countries in addition to India. Specifically, we propose how unified models can be built that can exploit the data from comparatively resource-rich languages of the same region. We propose strategies to unify different types of unexplored scripts, especially Perso-Arabic scripts and Indic scripts to build multilingual models for all the South Asian languages despite the script barrier. We also study how augmentation techniques like back-translation can be made useof to build unified models just using openly available raw data, to understand what levels of improvements can be expected for these Indic languages.

show abstract

Low Resource ASR: The Surprising Effectiveness of High Resource Transliteration

Cited by 21 publications

References 9 publications

MyMove: Facilitating Older Adults to Collect In-Situ Activity Labels on a Smartwatch with Speech

MyMove: Facilitating Older Adults to Collect In-Situ Activity Labels on a Smartwatch with Speech

Thinking and exploration into intellectualization of speech and medical acoustics

Unified NMT models for the Indian subcontinent, transcending script-barriers

Contact Info

Product

Resources

About