Behind-the-Scenes Challenges in Sino-Tibetan Low-Resource Language Technology Development
Abstract
AbstractDespite persistent efforts to develop low-resource language technologies (LRLTs), little attention has been paid to LRLT practitioners’ experiences, challenges, and opportunities. We employ a qualitative approach and conduct 15 semi-structured interviews with LRLT practitioners working on three Sino-Tibetan low-resource languages. Through our analysis, we identify not only technical challenges such as a lack of textual data and reliable language processing tools, but also linguistic, social, and financial challenges that cause and exacerbate them. Fortunately, we also find that participants are generally optimistic about LRLT development and mitigate these challenges by fostering and seeking support from diverse sources and leveraging large language models. We discuss the complex nature of the challenges in relation to prior literature and provide recommendations for practitioners, governments, and the broader NLP community.