INTENTIONQA

INTENTIONQA是由香港科技大学计算机科学与工程系创建的一个多选题问答数据集，专注于评估语言模型在电子商务中理解购买意图的能力。该数据集包含4,360个精心设计的问题，分为三个难度级别，通过自动化流程构建，确保在大规模电子商务平台上的可扩展性。数据集通过将人类注释的意图从FolkScope转换为问题，利用ASER事件知识图谱进行自动干扰项采样，以提高数据集质量。INTENTIONQA的应用领域主要集中在改进电子商务搜索服务，通过增强语言模型对购买意图的理解和利用，提升产品推荐等下游任务的效果。

INTENTIONQA is a multiple-choice question answering (QA) dataset developed by the Department of Computer Science and Engineering at the Hong Kong University of Science and Technology, focusing on evaluating language models' capability to comprehend purchase intentions in the e-commerce domain. This dataset contains 4,360 carefully designed questions categorized into three difficulty levels, and is constructed via an automated pipeline to ensure scalability across large-scale e-commerce platforms. Specifically, it converts human-annotated purchase intentions from FolkScope into question structures, and leverages the ASER event knowledge graph for automatic distractor sampling to enhance the dataset's quality. The main application scenarios of INTENTIONQA center on improving e-commerce search services, by strengthening language models' understanding of purchase intentions and utilizing such capabilities to elevate the performance of downstream tasks such as product recommendation.