ParaQA

ParaQA是一个针对知识图谱的单轮对话问答数据集，由波恩大学创建。该数据集包含5000个问题-答案对，每个问题至少有两个，最多有八个独特的改写答案。数据集通过半自动框架生成，利用了如反向翻译等高级改写技术。创建过程中，数据集继承了LC-QuAD的问题和VQuAnDa的答案表述，通过多步骤自动化处理生成多样化的答案。ParaQA的应用领域主要集中在提升单轮对话问答系统的性能，通过提供多样的答案表述来增强机器学习模型的表现。

ParaQA is a single-turn conversational question answering dataset for knowledge graphs, created by the University of Bonn. It comprises 5,000 question-answer pairs, with each question having at least two and up to eight distinct paraphrased answers. The dataset is generated via a semi-automatic framework that leverages advanced paraphrasing techniques such as back-translation. During its curation, the dataset inherits questions from LC-QuAD and answer formulations from VQuAnDa, generating diverse answers through multi-step automated processing. The primary application of ParaQA is to enhance the performance of single-turn conversational question answering systems by providing diverse answer paraphrases to improve machine learning model performance.