指南 3个标注团队操作手册:提升标注速度与质量

段落

Paragraphs标签在标注界面上显示文本段落。用于为NLP和NER项目标注对话记录。 Paragraphs标签期望的任务数据格式为如下对象数组: [{ $nameKey: "作者姓名", $textKey: "文本" }, ... ]

适用于以下数据类型:文本。

参数

参数 类型 默认值 描述
name string Name of the element
value string Data field containing the paragraph content
[valueType] json | url json Whether the data is stored directly in uploaded JSON data or needs to be loaded from a URL
[audioUrl] string Audio to sync phrases with
[sync] string Object name to sync with
[showPlayer] boolean false Whether to show audio player above the paragraphs. Ignored if sync object is audio
[saveTextResult] no | yes yes Whether to store labeled text along with the results. By default, doesn’t store text for valueType=url
[layout] none | dialogue none Whether to use a dialogue-style layout or not
[nameKey] string "author" The key field to use for name
[textKey] string "text" The key field to use for the text
[contextScroll] boolean false Turn on contextual scroll mode

结果参数

名称 类型 描述
value Object
value.start number index of paragraph where the region starts
value.end number index of paragraph where the region ends
value.startOffset number offset within start paragraph
value.endOffset number offset within end paragraph
[value.text] string text content of the region, can be skipped

示例JSON

{
  "value": {
    "start": 3,
    "end": 5,
    "startOffset": 2,
    "endOffset": 81,
    "paragraphlabels": ["Car"]
  }
}

示例

用于标注包含对话的文本段落区域的标注配置

<View>
  <Paragraphs name="dialogue-1" value="$dialogue" layout="dialogue" />
  <ParagraphLabels name="importance" toName="dialogue-1">
    <Label value="Important content"></Label>
    <Label value="Random talk"></Label>
  </ParagraphLabels>
</View>

示例

带音频的段落

<View>
  <Paragraphs audioUrl="$audio" value="$para" name="paragraphs"
              layout="dialogue" textKey="text" nameKey="author"
              showPlayer="true"
              />

  <Choices name="choices" toName="paragraphs" choice="multiple">
      <Choice value="Good quality"/>
      <Choice value="Fast speech"/>
  </Choices>
</View>

<!-- {"data": {
  "para": [
    {"text": "test 1", "author": "A", "start": 0.0, "end": 1.0},
    {"text": "test 2", "author": "B", "start": 1.0, "end": 2.0},
    {"text": "test 3", "author": "A", "start": 2.0, "end": 3.0}
  ],
  "audio": "/static/samples/game.wav"
}}
-->
专为各种规模的团队设计 版本比较