QUANT - Question Answering Benchmark Curator

Abstract

Question answering engines have become one of the most popular type of applications driven by Semantic Web technologies. Consequently, the provision of means to quantify the performance of current question answering approaches on current datasets has become ever more important. However, a large percentage of the queries found in popular question answering benchmarks cannot be executed on current versions of their reference dataset. There is a consequently a clear need to curate question answering benchmarks periodically. However, the manual alteration of question answering benchmarks is often error-prone. We alleviate this problem by presenting QUANT, a novel framework for the creation and curation of question answering benchmarks. QUANT supports the curation of benchmarks by generating smart edit suggestions for question-query pair and for the corresponding metadata. In addition, our framework supports the creation of new benchmark entries by providing predefined quality checks for queries. We evaluate QUANT on 653 questions obtained from QALD-1 to QALD-8 with 10 users. Our results show that our framework generates reliable suggestions and can reduce the curation effort for QA benchmarks by up to 91%.