CI-ToD
收藏资源简介:
CI-ToD是由哈尔滨工业大学社会计算与信息检索研究中心创建的一个面向任务导向对话系统的新型数据集,旨在解决对话系统中的不一致性问题。该数据集包含318条对话,通过人工标注确保每条对话的质量。数据集不仅标注了单一的不一致性标签,还提供了更细粒度的标签,如对话历史不一致性、用户查询不一致性和知识库不一致性,以帮助模型分析不一致性的来源。CI-ToD的应用领域主要集中在提高任务导向对话系统的一致性识别能力,解决现有模型在处理复杂对话时产生的不一致问题。
CI-ToD is a novel dataset for task-oriented dialogue systems, developed by the Social Computing and Information Retrieval Research Center of Harbin Institute of Technology, which aims to resolve inconsistency issues in dialogue systems. This dataset comprises 318 dialogues, with the quality of each dialogue guaranteed via manual annotation. Rather than only providing a single inconsistency label, the dataset offers more fine-grained tags including dialogue history inconsistency, user query inconsistency, and knowledge base inconsistency, to assist models in analyzing the sources of inconsistencies. The main application scenarios of CI-ToD focus on improving the consistency recognition capability of task-oriented dialogue systems and addressing the inconsistency problems arising when existing models handle complex dialogues.




