--- language: - code pretty_name: "Chinese ner dataseet" tags: - ner license: "bsd" task_categories: - token-classification --- 来源 https://github.com/liucongg/NLPDataSet * 从网上收集数据,将CMeEE数据集、IMCS21_task1数据集、CCKS2017_task2数据集、CCKS2018_task1数据集、CCKS2019_task1数据集、CLUENER2020数据集、MSRA数据集、NLPCC2018_task4数据集、CCFBDCI数据集、MMC数据集、WanChuang数据集、PeopleDairy1998数据集、PeopleDairy2004数据集、GAIIC2022_task2数据集、WeiBo数据集、ECommerce数据集、FinanceSina数据集、BoSon数据集、Resume数据集、Bank数据集、FNED数据集和DLNER数据集等22个数据集进行整理清洗,构建一个较完善的中文NER数据集。 * 数据集清洗时,仅进行了简单地规则清洗,并将格式进行了统一化,标签为“BIO”。 * 处理后数据集详细信息,见[数据集描述](https://zhuanlan.zhihu.com/p/529541521)。 * 数据集由[NJUST-TB](https://github.com/Swag-tb)一起整理。 * 由于部分数据包含嵌套实体的情况,所以转换成BIO标签时,长实体会覆盖短实体。 | 数据 | 原始数据/项目地址 | 样本个数 | 类别 | 原始数据描述 | | ------ | ------ | ------ | ------ | ------ | | CMeEE数据集 | [地址](http://www.cips-chip.org.cn/2021/CBLUE) | 20000条 | sym、dep、dru、pro、equ、dis、mic、ite和bod | 中文医疗信息处理挑战榜CBLUE中医学实体识别数据集 | | IMCS21_task1数据集 | [地址](http://www.fudan-disc.com/sharedtask/imcs21/index.html?spm=5176.12282016.0.0.140e6d92ypyW1r) | 98452条 | Operation、Drug_Category、Medical_Examination、Symptom和Drug | CCL2021第一届智能对话诊疗评测比赛命名实体识别数据集| | CCKS2017_task2数据集 | [地址](https://www.biendata.xyz/competition/CCKS2017_2/) | 2229条 | symp、dise、chec、body和cure | CCKS2017面向电子病历的命名实体识别数据集 | | CCKS2018_task1数据集 | [地址](https://www.biendata.xyz/competition/CCKS2018_1/) | 797条 | 症状和体征、检查和检验、治疗、疾病和诊断、身体部位 | CCKS2018面向中文电子病历的命名实体识别数据集 | | CCKS2019_task1数据集 | [地址](http://openkg.cn/dataset/yidu-s4k) | 1379条 | 解剖部位、手术、疾病和诊断、药物、实验室检验、影像检查 | CCKS2019面向中文电子病历的命名实体识别数据集 | | CLUENER2020数据集 | [地址](https://github.com/CLUEbenchmark/CLUENER2020) | 12091条 | game、organization、government、movie、name、book、company、scene、position和address | CLUENER2020数据集 | | MSRA数据集 | [地址](https://www.msra.cn/) | 48442条 | LOC、ORG和PER | MSRA微软亚洲研究院开源命名实体识别数据集 | | NLPCC2018_task4数据集 | [地址](http://tcci.ccf.org.cn/conference/2018/taskdata.php) | 21352条 | language、origin、theme、custom_destination、style、phone_num、destination、contact_name、age、singer、song、instrument、toplist、scene和emotion | 任务型对话系统数据数据集 | | CCFBDCI数据集 | [地址](https://www.datafountain.cn/competitions/510) |15723条 | LOC、GPE、ORG和PER | 中文命名实体识别算法鲁棒性评测数据集 | | MMC数据集 | [地址](https://tianchi.aliyun.com/competition/entrance/231687/information) | 3498条 | Level、Method、Disease、Drug、Frequency、Amount、Operation、Pathogenesis、Test_items、Anatomy、Symptom、Duration、Treatment、Test_Value、ADE、Class、Test和Reason | 瑞金医院MMC人工智能辅助构建知识图谱大赛数据集 | | WanChuang数据集 | [地址](https://tianchi.aliyun.com/competition/entrance/531827/introduction) | 1255条 | 药物剂型、疾病分组、人群、药品分组、中药功效、症状、疾病、药物成分、药物性味、食物分组、食物、证候和药品 | "万创杯”中医药天池大数据竞赛—智慧中医药应用创新挑战赛数据集 | | PeopleDairy1998数据集 | [地址]() | 27818条 | LOC、ORG和PER | 人民日报1998数据集 | | PeopleDairy2004数据集 | [地址]() | 286268条 | LOC、ORG、PER、T | 人民日报2004数据集 | | GAIIC2022_task2数据集 | [地址](https://www.heywhale.com/home/competition/620b34ed28270b0017b823ad/content/2) | 40000条 | 该比赛共有52种类别 | 商品标题实体识别数据集 | | WeiBo数据集 | [地址](https://github.com/hltcoe/golden-horse) | 1890条 | LOC.NAM、LOC.NOM、PER.NAM、ORG.NOM、ORG.NAM、GPE.NAM和PER.NOM | 社交媒体中文命名实体识别数据集 | | ECommerce数据集 | [地址](https://github.com/allanj/ner_incomplete_annotation) | 7998条 | MISC、XH、HPPX和HCCX | 面向电商的命名实体识别数据集 | | FinanceSina数据集 | [地址](https://github.com/jiesutd/LatticeLSTM) | 1579条 | LOC、GPE、ORG和PER | 新浪财经爬取中文命名实体识别数据集 | | BoSon数据集 | [地址](https://github.com/bosondata) | 2000条 | time、product_name、person_name、location、org_name和company_name | 玻森中文命名实体识别数据集 | | Resume数据集 | [地址](https://github.com/jiesutd/LatticeLSTM/tree/master/ResumeNER) | 4761条 | NAME、EDU、LOC、ORG、PRO、TITLE、CONT和RACE | 中国股市上市公司高管的简历 | | Bank数据集 | [地址](https://www.heywhale.com/mw/dataset/617969ec768f3b0017862990/file) | 10000条 | BANK、COMMENTS_ADJ、COMMENTS_N和PRODUCT | 银行借贷数据数据集 | | FNED数据集 | [地址](https://www.datafountain.cn/competitions/561/datasets) | 10500条 | LOC、GPE、ORG、EQU、TIME、FAC和PER | 高鲁棒性要求下的领域事件检测数据集 | | DLNER数据集 | [地址](https://github.com/lancopku/Chinese-Literature-NER-RE-Dataset) | 28897条 | Location、Thing、Abstract、Organization、Metric、Time、Physical、Person和Term | 语篇级命名实体识别数据集 | - 清洗及格式转换后的数据,下载链接如下:[百度云](https://pan.baidu.com/s/1VvbvWPv3eM4MXsv_nlDSSA) / 提取码:4sea - 注意:部分嵌套实体的数据,使用长实体覆盖了短实体,有嵌套实体需求的同学,请自行使用原始数据。