219 lines
5.6 KiB
Markdown
219 lines
5.6 KiB
Markdown
# WNUT2017
|
||
|
||
## 概述
|
||
|
||
WNUT2017 数据集收集了来自 Twitter 和 YouTube 等多个社交媒体平台的用户生成文本，专门用于命名实体识别（NER）任务，重点关注新兴和非常规实体。
|
||
|
||
## 任务描述
|
||
|
||
- **任务类型**：新兴实体命名实体识别（NER）
|
||
- **输入**：用户生成的社交媒体文本
|
||
- **输出**：带类型的已识别实体片段
|
||
- **领域**：社交媒体、新兴实体
|
||
|
||
## 主要特点
|
||
|
||
- 来自多个平台的用户生成文本
|
||
- 聚焦于新兴和非常规命名实体
|
||
- 包含六种实体类型，覆盖多样
|
||
- 非正式文本处理具有挑战性
|
||
- 源自 WNUT 2017 共享任务
|
||
|
||
## 评估说明
|
||
|
||
- 默认配置使用 **5-shot** 评估
|
||
- 评估指标：精确率（Precision）、召回率（Recall）、F1 分数（F1-Score）、准确率（Accuracy）
|
||
- 实体类型：CORPORATION（公司）、CREATIVE-WORK（创意作品）、GROUP（群体）、LOCATION（地点）、PERSON（人物）、PRODUCT（产品）
|
||
|
||
## 属性
|
||
|
||
| 属性 | 值 |
|
||
|----------|-------|
|
||
| **基准测试名称** | `wnut2017` |
|
||
| **数据集ID** | [extraordinarylab/wnut2017](https://modelscope.cn/datasets/extraordinarylab/wnut2017/summary) |
|
||
| **论文** | N/A |
|
||
| **标签** | `Knowledge`, `NER` |
|
||
| **指标** | `precision`, `recall`, `f1_score`, `accuracy` |
|
||
| **默认样本数** | 5-shot |
|
||
| **评估划分** | `test` |
|
||
| **训练划分** | `train` |
|
||
|
||
## 数据统计
|
||
|
||
| 指标 | 值 |
|
||
|--------|-------|
|
||
| 总样本数 | 1,287 |
|
||
| 提示词长度（平均） | 2665.65 字符 |
|
||
| 提示词长度（最小/最大） | 2570 / 3093 字符 |
|
||
|
||
## 样例示例
|
||
|
||
**子集**: `default`
|
||
|
||
```json
|
||
{
|
||
"input": [
|
||
{
|
||
"id": "f1e6b117",
|
||
"content": "Here are some examples of named entity recognition:\n\nInput:\n@paulwalk It 's the view from where I 'm living for two weeks . Empire State Building = ESB . Pretty bad storm here last evening .\n\nOutput:\n<response>@paulwalk It 's the view from wh ... [TRUNCATED] ... he most specific entity type.\n7. Ensure every opening tag has a matching closing tag.\n\nText to process:\n& gt ; * The soldier was killed when another avalanche hit an army barracks in the northern area of Sonmarg , said a military spokesman .\n"
|
||
}
|
||
],
|
||
"target": "<response>& gt ; * The soldier was killed when another avalanche hit an army barracks in the northern area of <location>Sonmarg</location> , said a military spokesman .</response>",
|
||
"id": 0,
|
||
"group_id": 0,
|
||
"metadata": {
|
||
"tokens": [
|
||
"&",
|
||
"gt",
|
||
";",
|
||
"*",
|
||
"The",
|
||
"soldier",
|
||
"was",
|
||
"killed",
|
||
"when",
|
||
"another",
|
||
"avalanche",
|
||
"hit",
|
||
"an",
|
||
"army",
|
||
"barracks",
|
||
"in",
|
||
"the",
|
||
"northern",
|
||
"area",
|
||
"of",
|
||
"Sonmarg",
|
||
",",
|
||
"said",
|
||
"a",
|
||
"military",
|
||
"spokesman",
|
||
"."
|
||
],
|
||
"ner_tags": [
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"B-LOCATION",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O",
|
||
"O"
|
||
]
|
||
}
|
||
}
|
||
```
|
||
|
||
*注：部分内容为显示目的已被截断。*
|
||
|
||
## 提示模板
|
||
|
||
**提示模板：**
|
||
```text
|
||
You are a named entity recognition system that identifies the following entity types:
|
||
{entities}
|
||
|
||
Process the provided text and mark all named entities with XML-style tags.
|
||
|
||
For example:
|
||
<person>John Smith</person> works at <organization>Google</organization> in <location>Mountain View</location>.
|
||
|
||
Available entity tags: {entity_list}
|
||
|
||
INSTRUCTIONS:
|
||
1. Wrap your entire response in <response>...</response> tags.
|
||
2. Inside these tags, include the original text with entity tags inserted.
|
||
3. Do not change the original text in any way (preserve spacing, punctuation, case, etc.).
|
||
4. Tag ALL entities you can identify using the exact tag names provided.
|
||
5. Do not include explanations, just the tagged text.
|
||
6. If entity spans overlap, choose the most specific entity type.
|
||
7. Ensure every opening tag has a matching closing tag.
|
||
|
||
Text to process:
|
||
{text}
|
||
|
||
```
|
||
|
||
<details>
|
||
<summary>少样本（Few-shot）模板</summary>
|
||
|
||
```text
|
||
Here are some examples of named entity recognition:
|
||
|
||
{fewshot}
|
||
|
||
You are a named entity recognition system that identifies the following entity types:
|
||
{entities}
|
||
|
||
Process the provided text and mark all named entities with XML-style tags.
|
||
|
||
For example:
|
||
<person>John Smith</person> works at <organization>Google</organization> in <location>Mountain View</location>.
|
||
|
||
Available entity tags: {entity_list}
|
||
|
||
INSTRUCTIONS:
|
||
1. Wrap your entire response in <response>...</response> tags.
|
||
2. Inside these tags, include the original text with entity tags inserted.
|
||
3. Do not change the original text in any way (preserve spacing, punctuation, case, etc.).
|
||
4. Tag ALL entities you can identify using the exact tag names provided.
|
||
5. Do not include explanations, just the tagged text.
|
||
6. If entity spans overlap, choose the most specific entity type.
|
||
7. Ensure every opening tag has a matching closing tag.
|
||
|
||
Text to process:
|
||
{text}
|
||
|
||
```
|
||
|
||
</details>
|
||
|
||
## 使用方法
|
||
|
||
### 使用命令行（CLI）
|
||
|
||
```bash
|
||
evalscope eval \
|
||
--model YOUR_MODEL \
|
||
--api-url OPENAI_API_COMPAT_URL \
|
||
--api-key EMPTY_TOKEN \
|
||
--datasets wnut2017 \
|
||
--limit 10 # 正式评估时请删除此行
|
||
```
|
||
|
||
### 使用 Python
|
||
|
||
```python
|
||
from evalscope import run_task
|
||
from evalscope.config import TaskConfig
|
||
|
||
task_cfg = TaskConfig(
|
||
model='YOUR_MODEL',
|
||
api_url='OPENAI_API_COMPAT_URL',
|
||
api_key='EMPTY_TOKEN',
|
||
datasets=['wnut2017'],
|
||
limit=10, # 正式评估时请删除此行
|
||
)
|
||
|
||
run_task(task_cfg=task_cfg)
|
||
``` |