情报科学 ›› 2024, Vol. 42 ›› Issue (3): 118-128.

• 业务研究 • 上一篇    下一篇

中国手语基本手势语义知识组织研究

  

  • 出版日期:2024-03-05 发布日期:2024-06-08

  • Online:2024-03-05 Published:2024-06-08

摘要:

【目的/意义】为了挖掘手语语料蕴含的丰富语义知识,提高手语资源的利用率,提出一种手语基本手势语
义知识提取与知识组织方法。【方法/过程】首先,建立手语词汇多模态特征标注体系,实施标注方案;其次,通过计
算手势相似性,抽取基本手势,形成基本手势--手语词汇簇;再次,通过对基本手势——手语词汇簇进行多层次聚
类,提取基本手势语义项;最后,建立基本手势基本属性与语义知识表示模型,并进行存储。【结果/结论】基于国家
手语词汇语料库的上海手语词汇语料,抽取基本手势1474个,确定义项4034项,并开发出基本手势语义知识系统。
【创新/局限】本研究提出了一种基于数据驱动的方法,用于提取和组织手语基本手势语义知识,并开发了相应的系
统平台,为手语语义分析和相关领域提供了宝贵的资源基础。本文主要分析是单语素手势的语义知识,未涉及多
语素手势、分类词谓语等复杂结构的语义分析。

Abstract:

【Purpose/significance】 In order to explore the rich semantic knowledge embedded in sign language corpus and improve the
utilization of sign language resources, this study proposes a method for extracting and organizing semantic knowledge from Chinese
sign language basic gestures.【Method/process】 Firstly, a multimodal feature annotation system for sign language vocabulary is estab⁃
lished and implemented through an annotation scheme. Secondly, basic gestures are extracted by calculating gesture similarity, form⁃
ing clusters of basic gestures and sign language vocabulary. Thirdly, hierarchical clustering is performed on the clusters of basic ges⁃
tures and sign language vocabulary to extract semantic items of basic gestures. Finally, a representation model for basic gesture attri⁃
butes and semantic knowledge is established and stored.【Result/conclusion】 Based on the Shanghai sign language vocabulary corpus
from the National Sign Language Lexicon, 1474 basic gestures are extracted and 4034 semantic items are identified, and a system for
basic gesture semantic knowledge is developed.【Innovation/limitation】 This study proposes a data-driven method for extracting and
organizing the semantic knowledge of basic gestures, and develops a corresponding system platform, providing valuable resource foun⁃
dation for sign language semantic analysis and related fields. The focus of this paper is the semantic knowledge of moneme gestures,
without involving the semantic analysis of complex structures such as multi-morpheme gestures and classifying predicates.