ConditionalQA
收藏资源简介:
ConditionalQA是一个复杂的阅读理解数据集,专注于具有条件答案的问题。该数据集由卡内基梅隆大学计算机科学学院创建,包含3617个问题,这些问题不仅包括提取式问题、是/否问题,还包括多答案问题和无法回答的问题。数据集的特点在于其文档结构复杂,需要模型进行多跳逻辑推理以找到正确答案。此外,数据集中的问题在提出时并不知道答案,这模拟了真实世界中信息寻求的过程。ConditionalQA的应用领域主要在于推动对长文档中复杂问题的回答研究,特别是在理解和预测条件答案方面。
ConditionalQA is a complex reading comprehension dataset focused on questions with conditional answers. It was developed by the School of Computer Science at Carnegie Mellon University, and encompasses 3,617 questions covering extractive questions, yes/no questions, multi-answer questions, and unanswerable questions. A defining characteristic of this dataset is its complex document structure, which requires models to perform multi-hop logical reasoning to identify the correct answers. Furthermore, the questions in the dataset are posed without prior knowledge of their answers, which simulates the real-world information-seeking process. The primary application scope of ConditionalQA lies in advancing research on complex question answering over long documents, particularly in the understanding and prediction of conditional answers.

- 1ConditionalQA: A Complex Reading Comprehension Dataset with Conditional Answers卡内基梅隆大学计算机科学学院 · 2021年



