Recent years have witnessed the adoption of parameter-efficient adapters in pre-trained language models for natural language processing. Yet, their application in speech processing remains less studied. In this work, we explore the adapters for low-resource speech recognition, introducing a novel technique - ConvAdapt into pre-trained speech models. We investigate various aspects such as data requirements, transfer learning within adapters, and scaling of feed-forward layers in adapters. Our findings reveal that bottleneck adapters offer competitiveness with full fine-tuning with at least 10 hours of data, but they are not as effective in few-shot learning scenarios. Notably, ConvAdapt demonstrates improved performance in such cases. In addition, transfer learning in adapters shows promise, necessitating research in related languages. Furthermore, employing larger speech models for adapter-tuning surpasses fine-tuning with ample data, potentially due to reduced overfitting than fine-tuning.

Multilingualism is deeply rooted in the sociopolitical history of Thailand. Some minority language communities entered the Thai territory a few decades ago, while the families of some other minority speakers have been living in Thailand since at least several generations. The authors of this article address the question how Akha, Dara-ang, Karen, Khamu, Mlabri and Urak Lawoi’ language speakers perceive the current situation of their language and whether they see the need for the development of digital tools for documentation, revitalization and daily use of their languages. The objective is complemented by a discussion on the feasibility of development of such tools for some of the above mentioned languages and the motivation of their speakers to participate in this process. Furthermore, this article highlights the challenges associated with developing digital tools for these low-resource languages and outlines the standards researchers must adhere to in conceptualizing the development of such tools, collecting data, and engaging with the language communities throughout the collaborative process.