摘要
手语语料库是支撑手语语言学研究、手语规范化、手语识别与合成技术发展及信息无障碍建设的重要基础设施。国际手语语料库建设始于20世纪90年代,经过20余年的发展,特别是伴随着计算机技术的进步,在语料规模、标注规范和数据组织等方面已取得显著进展,但仍面临数据模态单一、标注标准不统一等挑战。系统梳理国际大规模手语语料库的类型、规模与建设方法,重点发掘数据采集、标注体系与数据组织等核心环节,可为对比分析国内手语语料库建设现状与不足提供参考借鉴。当前,中国手语语料库建设尚处于初级阶段,存在规模小、类型单一、标准缺失等问题。建议:融合大语言模型与跨模态对齐技术,提升数据采集效率与规模,整合多层级数据库构建多样化语料库体系;建立统一编码规范与标注标准,开发中国手语编码字符集,推动标注自动化转型;构建多方协作机制,鼓励聋人社群参与语料建设,建立共建共享平台以实现可持续发展。
Abstract
Sign language corpora constitute essential infrastructure for advancing sign language linguistic research, sign language standardization, sign language recognition and synthesis technologies, and information accessibility. The development of sign language corpora worldwide began in the 1990s. Over the past two decades, particularly with advances in computer technology, significant progress has been made in corpus size, annotation conventions, and data organization. Nevertheless, challenges remain, including limited diversity in data modalities and a lack of unified annotation standards. This paper systematically reviews the types, scales, and construction methodologies of large-scale sign language corpora worldwide, with particular emphasis on key aspects such as data collection, annotation systems, and data organization. By examining international practices and experiences, this study provides a basis for identifying current limitations of sign language corpus development in China and exploring potential directions for improvement. At present, sign language corpus development in China remains at an early stage, characterized by limited corpus size, insufficient diversity in corpus types, and a lack of standardized frameworks. To address these challenges, this paper proposes several recommendations. First, large language models and cross-modal alignment technologies should be integrated to enhance the efficiency and scale of data collection, while multi-level databases should be incorporated to establish a diversified sign language corpus system. Second, unified coding conventions and annotation standards should be established, alongside the development of a dedicated character encoding system for Chinese Sign Language, to facilitate the transition towards automated annotation. Third, multi-stakeholder collaboration mechanisms should be established to encourage the active participation of Deaf communities in corpus development, supported by collaborative platforms for corpus creation and sharing to ensure long-term sustainability.
关键词
大规模手语语料库 /
多模态语言资源 /
数据采集与标注 /
建设规范
Key words
large-scale sign corpus /
multimodal linguistic resources /
data collection and annotation /
development protocols
倪 兰, 韩雨欣, 温育梓.
国际大规模手语语料库建设发展及启示[J]. 语言战略研究. 2026, 11(6): 74-85 https://doi.org/10.19689/j.cnki.cn10-1361/h.20260607
Ni Lan, Han Yuxin and Wen Yuzi.
International Developments in Large-Scale Sign Language Corpus Construction and Implications[J]. Chinese Journal of Language Policy and Planning. 2026, 11(6): 74-85 https://doi.org/10.19689/j.cnki.cn10-1361/h.20260607
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
基金
国家社会科学基金一般项目“基于多模态语料库的中国手语否定范畴研究”(20BYY177)。