ANTIQUE
收藏资源简介:
ANTIQUE是一个非事实性问答数据集,由马萨诸塞大学阿默斯特分校智能信息检索中心创建,包含2,626个开放领域的非事实性问题,这些问题来自Yahoo! Answers社区问答服务。数据集大小为34,011个手动相关性标注,数据来源于真实用户提问。创建过程中,通过多轮众包和自动及手动质量检查收集了四级相关性判断。ANTIQUE的应用领域是提高现代信息检索系统中非事实性问题的答案段落检索质量,旨在解决现有数据集在相关性判断和问题多样性方面的不足。
ANTIQUE is a non-factoid question answering dataset created by the Center for Intelligent Information Retrieval at the University of Massachusetts Amherst. It consists of 2,626 open-domain non-factoid questions sourced from the Yahoo! Answers community question answering service. The dataset includes 34,011 manually annotated relevance judgments derived from real user queries. During its development, four-level relevance judgments were collected through multi-round crowdsourcing combined with both automatic and manual quality checks. The target application scenario of ANTIQUE is to improve the quality of answer passage retrieval for non-factoid questions in modern information retrieval systems, aiming to address the shortcomings of existing datasets in terms of relevance judgment and question diversity.

- 1ANTIQUE: A Non-Factoid Question Answering Benchmark马萨诸塞大学阿默斯特分校智能信息检索中心 · 2019年



