ProCQA
收藏资源简介:
ProCQA,由北京航空航天大学构建的一个大规模编程问答数据集,包含约500万个问答对,覆盖Python、Java、JavaScript等11种不同的编程语言,内容涉及算法、框架、库的使用等多个知识领域。该数据集源自StackOverflow社区,研究人员通过爬虫技术获取,采用了严格的规则过滤策略,如过滤过短或过长的问答,只保留被提问者接受的答案等,确保数据的质量和公平性。ProCQA中的问答对是自然结构化的混合模态对,即文本和代码在问答字段中交织在一起,为模型提供了自然监督信号,以便于对齐两种模态。ProCQA的应用领域广泛,不仅可以用作评估基准,也可以用作预训练语料库,为代码检索和问答任务提供了宝贵的资源。
ProCQA is a large-scale programming question answering (QA) dataset constructed by Beihang University. It contains approximately 5 million QA pairs, covering 11 programming languages including Python, Java, JavaScript and others, involving multiple knowledge domains such as algorithms, framework and library usage. This dataset is sourced from the StackOverflow community, and was collected by researchers via web crawling technologies. Strict rule-based filtering strategies were applied, such as filtering out overly short or long QA pairs and only retaining answers accepted by questioners, to ensure the quality and fairness of the dataset. The QA pairs in ProCQA are naturally structured mixed-modal pairs, where text and code are interleaved in the question and answer fields, providing natural supervision signals for models to align the two modalities. ProCQA has broad application scenarios: it can not only be used as an evaluation benchmark, but also serve as a pre-training corpus, providing valuable resources for code retrieval and QA tasks.
ProCQA 数据集概述
数据集描述
- 名称: ProCQA
- 类型: 社区基础的编程问答数据集
- 获取方式: 通过百度网盘下载,链接为 https://pan.baidu.com/s/1cOSlbPZe8ekn1JwVbjeoSw?pwd=xxd2




