构建你的第一个命名实体识别标注任务

命名实体识别（NER）是最常见的 NLP 任务之一。在本教程中，你将学习如何创建一个完整的 NER 标注界面，包含文本高亮、键盘快捷键和实体类型选择。

我们要构建的内容

完成本教程后，你将拥有一个标注界面，标注者可以：

通过点击和拖动来高亮文本片段
分配实体类型（人物、组织、地点等）
使用键盘快捷键加速标注
编辑或删除已有标注

前置条件

已安装 Potato（pip install potato-annotation）
基本的 YAML 知识
待标注的示例文本数据

第1步：配置标注方案

创建 config.yaml 文件：

yaml

annotation_task_name: "Named Entity Recognition"
 
data_files:
  - data/sentences.json
 
item_properties:
  id_key: id
  text_key: text
 
# Enable span annotation
annotation_schemes:
  - annotation_type: span
    name: entities
    description: "Highlight and label named entities in the text"
    labels:
      - name: PER
        description: "Person names"
        color: "#FF6B6B"
        keyboard_shortcut: "p"
      - name: ORG
        description: "Organizations"
        color: "#4ECDC4"
        keyboard_shortcut: "o"
      - name: LOC
        description: "Locations"
        color: "#45B7D1"
        keyboard_shortcut: "l"
      - name: DATE
        description: "Dates and times"
        color: "#96CEB4"
        keyboard_shortcut: "d"
      - name: MISC
        description: "Miscellaneous entities"
        color: "#FFEAA7"
        keyboard_shortcut: "m"
    min_spans: 0  # Allow sentences with no entities

第2步：准备数据

创建 data/sentences.json，包含你的文本数据：

json

{"id": "1", "text": "Apple Inc. announced that CEO Tim Cook will visit Paris next Tuesday."}
{"id": "2", "text": "The United Nations headquarters in New York hosted delegates from Japan."}
{"id": "3", "text": "Dr. Sarah Johnson published her research at Stanford University in March 2024."}

第3步：添加标注指南

用清晰的指南帮助你的标注者：

yaml

# Add to config.yaml
annotation_guidelines:
  title: "NER Annotation Guidelines"
  content: |
    ## Entity Types
 
    **PER (Person)**: Names of people, including fictional characters
    - Examples: "John Smith", "Dr. Johnson", "Batman"
 
    **ORG (Organization)**: Companies, institutions, agencies
    - Examples: "Apple Inc.", "United Nations", "Stanford University"
 
    **LOC (Location)**: Places, including countries, cities, landmarks
    - Examples: "Paris", "New York", "Mount Everest"
 
    **DATE**: Dates, times, and temporal expressions
    - Examples: "Tuesday", "March 2024", "next week"
 
    **MISC**: Other named entities not fitting above categories
    - Examples: "Nobel Prize", "iPhone", "COVID-19"
 
    ## Annotation Rules
    1. Include titles (Dr., Mr.) with person names
    2. For nested entities, annotate the largest meaningful span
    3. Don't include articles (the, a) in entity spans

第4步：开始标注

启动你的 NER 任务：

bash

potato start config.yaml

标注工作流程

选择文本：点击并拖动以高亮一个片段
选择实体类型：点击标签按钮或使用键盘快捷键
编辑标注：点击已有的 span 进行修改或删除
提交：完成后按 Enter 或点击提交

第5步：查看输出

标注结果以 JSONL 格式保存：

json

{
  "id": "1",
  "text": "Apple Inc. announced that CEO Tim Cook will visit Paris next Tuesday.",
  "annotations": {
    "entities": [
      {"start": 0, "end": 10, "label": "ORG", "text": "Apple Inc."},
      {"start": 30, "end": 38, "label": "PER", "text": "Tim Cook"},
      {"start": 50, "end": 55, "label": "LOC", "text": "Paris"},
      {"start": 61, "end": 73, "label": "DATE", "text": "next Tuesday"}
    ]
  }
}

更好的 NER 标注技巧

一致的指南：清晰的规则减少分歧
训练示例：在标注者开始前展示边缘案例
定期校准：团队讨论困难案例
衡量一致性：使用标注者间一致性来识别问题

下一步

添加训练阶段以引导标注者
设置多标注者以实现冗余
导出为 Hugging Face 格式用于模型训练

需要帮助？查看我们的 span 标注文档获取更多详情。