ACE 2005 corpus preprocessing for Event Extraction task
-
Prepare ACE 2005 dataset.
(Download: https://catalog.ldc.upenn.edu/LDC2006T06. Note that ACE 2005 dataset is not free.)
-
Install the packages.
pip install beautifulsoup4 nltk tqdm -
Choose your data_list，not given above. (train/dev/test)
Run:
sudo python main.py --data=./data/ace_2005/Chinese- Then you can get the parsed data in
output directory.
I follow the json format described in nlpcl-lab/ace2005-preprocessing [github] repository like the bellow sample. But currently only sentence, event-mentions, entity-mentions, others information such as dependency tree, pos_tags, etc. will be added later. The data division method (data_list.csv) is selected randomly during the experiment, it does not belong to the authoritative division method of ED task.
If you want to know event types and arguments in detail, read this document (ACE 2005 event guidelines).
sample.json
[
{
"sentence": "两个星期来，藤森曾亲自带队搜捕 前情报顾问蒙特西诺斯，迄今蒙特西诺斯仍未落网",
"golden-event-mentions": [
{
"arguments": [
{
"start": 29,
"end": 34,
"entity-type": "PER:Individual",
"text": "蒙特西诺斯",
"role": "Person"
},
{
"start": 0,
"end": 4,
"entity-type": "TIM:time",
"text": "两个星期",
"role": "Time"
}
],
"trigger": {
"start": 36,
"end": 38,
"text": "落网"
},
"event_type": "Justice:Arrest-Jail"
}
],
"golden-entity-mentions": [
{
"start": 16,
"entity-type": "PER:Individual",
"text": "前情报顾问",
"end": 21,
"phrase-type": "NOM"
},
{
"start": 21,
"entity-type": "PER:Individual",
"text": "蒙特西诺斯",
"end": 26,
"phrase-type": "NAM"
},
{
"start": 29,
"entity-type": "PER:Individual",
"text": "蒙特西诺斯",
"end": 34,
"phrase-type": "NAM"
},
{
"start": 6,
"entity-type": "PER:Individual",
"text": "藤森",
"end": 8,
"phrase-type": "NAM"
},
{
"start": 0,
"entity-type": "TIM:time",
"text": "两个星期",
"end": 4,
"phrase-type": "TIM"
},
{
"start": 27,
"entity-type": "TIM:time",
"text": "迄今",
"end": 29,
"phrase-type": "TIM"
}
]
},
]- nlpcl-lab's ace2005-preprocessing repository, [github]
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent