CONDAMBIGQA
收藏资源简介:
CONDAMBIGQA是一个包含200个模糊问题的基准和数据集,由岭南大学数据科学学院创建。该数据集通过检索-based的注释策略,使用检索到的Wikipedia片段来识别给定查询的可能解释作为其条件和注释答案。这种策略最小化了由于注释者不同知识水平而引入的人类偏见。CONDAMBIGQA旨在解决问答系统中的歧义问题,通过明确条件来系统地解决模糊性,确保模型的回答与用户的期望更紧密地对齐。
CONDAMBIGQA is a benchmark and dataset consisting of 200 ambiguous questions, created by the School of Data Science at Lingnan University. This dataset adopts a retrieval-based annotation strategy, using retrieved Wikipedia passages to identify possible interpretations of a given query as its conditions and annotated answers. This strategy minimizes human bias introduced by the varying knowledge levels of annotators. CONDAMBIGQA aims to address ambiguity issues in question answering systems, systematically resolving ambiguity through explicit condition specification to ensure that model responses align more closely with user expectations.

- 1CondAmbigQA: A Benchmark and Dataset for Conditional Ambiguous Question Answering岭南大学数据科学学院, 香港特别行政区 · 2025年



