Annotated Corpus of Reference Resolution for Interpreting Common Grounding

本数据集名为‘Annotated Corpus of Reference Resolution for Interpreting Common Grounding’，由东京大学创建，旨在通过参考解析研究共同基础的中间过程。数据集包含5,191个对话中的40,172个指称表达，每个表达都有至少三次的指称解释判断。数据集的创建过程涉及半自动标注指称表达和通过众包收集指称识别的判断。该数据集主要应用于对话系统中共同基础的解释和分析，旨在解决自然语言对话中复杂的不确定性和歧义问题。

This dataset, named *Annotated Corpus of Reference Resolution for Interpreting Common Grounding*, was developed by The University of Tokyo. It aims to explore the intermediate processes of common grounding via reference resolution. The corpus contains 40,172 referential expressions extracted from 5,191 dialogues, with at least three referential interpretation judgments for each expression. The construction of this dataset involved semi-automatically annotating referential expressions and collecting referential identification judgments via crowdsourcing. This corpus is primarily applied to the interpretation and analysis of common grounding in conversational systems, with the objective of resolving complex uncertainties and ambiguities in natural language dialogues.