KoSLA Corpus

KoSLA Corpus是由韩国延世大学创建的一个多模态手语增强语料库，专注于医院场景下的聋人沟通需求。该数据集包含约40,558条记录，通过同义词替换等数据增强技术生成，以提高翻译模型的效率和可用数据量，同时保持手语的语法和语义结构。KoSLA Corpus不仅包含手势信号，还考虑了面部表情和身体动作等非手势信号，以及象征性特征，旨在构建一个能够准确传达手语意义的机器翻译系统。该数据集的应用领域主要集中在医疗环境中，帮助医生和聋人进行有效沟通，同时也支持手语学习者的教育需求。

The KoSLA Corpus is a multimodal sign language augmented corpus developed by Yonsei University of South Korea, targeting the communication requirements of deaf individuals within hospital scenarios. Comprising approximately 40,558 records, this corpus was generated via data augmentation techniques including synonym replacement, with the goals of enhancing the efficiency and scaling up the available training data for translation models while retaining the grammatical and semantic structures of sign language. In addition to gestural signals, the KoSLA Corpus incorporates non-gestural signals such as facial expressions and body movements alongside symbolic features, with the ultimate objective of constructing a machine translation system capable of accurately conveying the semantic meaning of sign language. The primary application domains of this dataset lie in medical settings, where it facilitates effective communication between medical practitioners and deaf people while also supporting the educational needs of sign language learners.