ERP生态社区人工问答记录数据集

“金蝶云•苍穹”开放生态社区面向不同场景、不同需求的人员提供了ERP人工问答服务和ERP智能问答服务。当前，ERP人工问答服务收集并保存了大量问答记录，这些问答记录经过脱敏和去噪声后形成ERP人工问答数据集，不仅可以作为智能问答的数据参考，而且可以作为ERP智能问答机器人的训练数据，避免因缺少训练数据而造成冷启动问题。ERP生态社区人工问答记录数据集包含了超过300000条人工问答数据。数据是2021年12月18日到2022年3月9日期间，由金蝶软件（中国）有限公司生态社区日志系统采集金蝶技术人员与在线用户的人工问答交互记录后，经过电子科技大学研究人员进行数据去噪声、数据筛选和数据存储得到的。采集方式是金蝶公司自动日志采集，日志系统在人工问答服务高峰和低峰24小时不间断采集，数据类型是字符串文本，一个会话记录包含多次问和答的中文语句，共计26.5MB。

The Kingdee Cloud Galaxy Open Ecosystem Community provides ERP manual Q&A services and ERP intelligent Q&A services for users with diverse scenarios and varying needs. Currently, the ERP manual Q&A service has collected and stored a vast number of Q&A records. After undergoing data desensitization and noise reduction, these records form the ERP manual Q&A dataset, which can serve not only as a data reference for intelligent Q&A systems but also as training data for ERP intelligent Q&A robots, thus avoiding the cold start problem caused by insufficient training data. The ERP manual Q&A record dataset of the ecosystem community contains over 300,000 manual Q&A entries. Collected from December 18, 2021 to March 9, 2022, the dataset was captured by the ecological community log system of Kingdee Software (China) Co., Ltd., which recorded the manual Q&A interactions between Kingdee technical personnel and online users. It was subsequently processed by researchers from the University of Electronic Science and Technology of China (UESTC) through noise reduction, data filtering and data storage. The collection adopts automated log collection by Kingdee: the log system collects data 24 hours a day, 7 days a week during both peak and off-peak hours of the manual Q&A service. The data type is string text, where a single session record includes multiple Chinese question-and-answer utterances, with a total size of 26.5 MB.

搜集汇总

背景与挑战

背景概述

该数据集包含超过30万条金蝶云•苍穹ERP生态社区的人工问答记录，数据采集于2021年12月至2022年3月，经过脱敏和去噪处理，可用于智能问答系统的训练和参考。数据集由金蝶软件（中国）有限公司和电子科技大学合作整理，总大小为26.5MB，包含2个文件。