简体中文 | English
A Deep Learning Based Knowledge Extraction Toolkit for Knowledge Base Population
DeepKE is a knowledge extraction toolkit supporting **low-resource** and **document-level** scenarios. It provides three functions based **PyTorch**, including **Named Entity Recognition**, **Relation Extraciton** and **Attribute Extraction**.
## Online Demo
[demo](https://deepke.openkg.cn)
### Prediction
There is a demonstration of prediction.
## Model Framework
Figure 1: The framework of DeepKE
- DeepKE contains three modules for **named entity recognition**, **relation extraction** and **attribute extraction**, the three tasks respectively.
- Each module has its own submodules. For example, there are **standard**, **document-level** and **few-shot** submodules in the attribute extraction modular.
- Each submodule compose of three parts: a **collection of tools**, which can function as tokenizer, dataloader, preprocessor and the like, a **encoder** and a part for **training and prediction**
## Quickstart
Take the fully supervised attribute extraction for example.
1. Download basic codes `git clone https://github.com/zjunlp/DeepKE.git `
2. Create a virtual environment (recommend `anaconda`) `conda create -n deepke python=3.8`
3. Enter the environment `conda activate deepke`
4. Install dependent packages
- If use deepke directly: `pip install deepke`
- If modify source codes before usage:
run `python setup.py install` firstly,
after modification, run `python setup.py develop`
5. Enter the corresponding directory `cd DeepKE/example/re/standard`
6. Train `python run.py` (Parameters for training can be changed in the `conf` folder)
7. Predict `python predict.py`(Parameters for prediction can be changed in the `conf` folder)
### Requirements
> python == 3.8
- torch == 1.5
- hydra-core == 1.0.6
- tensorboard == 2.4.1
- matplotlib == 3.4.1
- transformers == 3.4.0
- jieba == 0.42.1
- scikit-learn == 0.24.1
- pytorch-transformers == 1.2.0
- seqeval == 1.2.2
- tqdm == 4.60.0
- opt-einsum==3.3.0
- ujson
### Introduction of Three Functions
#### 1. Named Entity Recognition
- Named entity recognition seeks to locate and classify named entities mentioned in unstructured text into pre-defined categories such as person names, organizations, locations, organizations, etc.
- The data is stored in `.txt` files. Some instances as following:
| Sentence | Person | Location | Organization |
| :----------------------------------------------------------: | :------------------------: | :------------: | :----------------------------: |
| 本报北京9月4日讯记者杨涌报道：部分省区人民日报宣传发行工作座谈会9月3日在4日在京举行。 | 杨涌 | 北京 | 人民日报 |
| 《红楼梦》是中央电视台和中国电视剧制作中心根据中国古典文学名著《红楼梦》摄制于1987年的一部古装连续剧，由王扶林导演，周汝昌、王蒙、周岭等多位红学家参与制作。 | 王扶林，周汝昌，王蒙，周岭 | 中国 | 中央电视台，中国电视剧制作中心 |
| 秦始皇兵马俑位于陕西省西安市，1961年被国务院公布为第一批全国重点文物保护单位，是世界八大奇迹之一。 | 秦始皇 | 陕西省，西安市 | 国务院 |
- Read the detailed process in specific README
- **[STANDARD (Fully Supervised)](https://github.com/zjunlp/deepke/blob/test_new_deepke/example/ner/standard)**
- The standard module is implemented by the pretrained model *BERT*.
- Enter `DeepKE/example/ner/standard`.
- The dataset and parameters can be customized in the `data` folder and `conf` folder respectively.
- **Train**: `python run.py`
- **Predict**: `python predict.py`
- **[FEW-SHOT](https://github.com/zjunlp/DeepKE/tree/test_new_deepke/example/ner/few-shot)**
- This module is in the low-resouce scenario.
- Enter `DeepKE/example/ner/few-shot`.
- The directory where the model is loaded and saved and the configuration parameters can be cusomized in the `conf` folder.
- **Train with *CoNLL-2003***: `python run.py`
- **Train in the few-shot scenario**: `python run.py +train=few_shot`. Users can modify `load_path` in `conf/train/few_shot.yaml` with the use of existing loaded model.
- **Predict**: add `- predict` to `conf/config.yaml`, modify `loda_path` as the model path and `write_path` as the path where the predicted results are saved in `conf/predict.yaml`, and then run `python predict.py`
#### 2. Relation Extraction
- Relationship extraction is the task of extracting semantic relations between entities from a unstructured text.
- The data is stored in `.csv` files. Some instances as following:
| Sentence | Relation | Head | Head_offset | Tail | Tail_offset |
| :----------------------------------------------------: | :------: | :--------: | :---------: | :--------: | :---------: |
| 《岳父也是爹》是王军执导的电视剧，由马恩然、范明主演。 | 导演 | 岳父也是爹 | 1 | 王军 | 8 |
| 《九玄珠》是在纵横中文网连载的一部小说，作者是龙马。 | 连载网站 | 九玄珠 | 1 | 纵横中文网 | 7 |
| 提起杭州的美景，西湖总是第一个映入脑海的词语。 | 所在城市 | 西湖 | 8 | 杭州 | 2 |
- Read the detailed process in specific README
- **[STANDARD (Fully Supervised)](https://github.com/zjunlp/deepke/blob/test_new_deepke/example/re/standard)**
- The standard module is implemented by common deep learning models, including CNN, RNN, Capsule, GCN, Transformer and the pretrained model.
- Enter the `DeepKE/example/re/standard` folder.
- The dataset and parameters can be customized in the `data` folder and `conf` folder respectively.
- **Train**: `python run.py`
- **Predict**: `python predict.py`
- **[FEW-SHOT](https://github.com/zjunlp/deepke/blob/test_new_deepke/example/re/few-shot)**
- This module is in the low-resouce scenario.
- Enter `DeepKE/example/re/few-shot` .
- **Train**: `python run.py`
Start with the model trained last time: modify `train_from_saved_model` in `conf/train.yaml`as the path where the model trained last time was saved.
And the path saving logs generated in training can be customized by `log_dir`.
- **Predict**: `python predict.py`
- **[DOCUMENT](https://github.com/zjunlp/deepke/blob/test_new_deepke/example/re/document)**
- Download the model `train_distant.json` from [*Google Drive*](https://drive.google.com/drive/folders/1c5-0YwnoJx8NS6CV2f-NoTHR__BdkNqw) to `data/`.
- Enter `DeepKE/example/re/document` .
- **Train**: `python run.py`
Start with the model trained last time: modify `train_from_saved_model` in `conf/train.yaml`as the path where the model trained last time was saved.
And the path saving logs generated in training can be customized by `log_dir`.
- **Predict**: `python predict.py`
#### 3. Attribute Extraction
- Attribute extraction is to extract attributes for entities in a unstructed text.
- The data is stored in `.csv` files. Some instances as following:
| Sentence | Att | Ent | Ent_offset | Val | Val_offset |
| :----------------------------------------------------------: | :------: | :------: | :--------: | :-----------: | :--------: |
| 张冬梅，女，汉族，1968年2月生，河南淇县人 | 民族 | 张冬梅 | 0 | 汉族 | 6 |
| 杨缨，字绵公，号钓溪，松溪县人，祖籍将乐，是北宋理学家杨时的七世孙 | 朝代 | 杨缨 | 0 | 北宋 | 22 |
| 2014年10月1日许鞍华执导的电影《黄金时代》上映 | 上映时间 | 黄金时代 | 19 | 2014年10月1日 | 0 |
- Read the detailed process in specific README
- **[STANDARD (Fully Supervised)](https://github.com/zjunlp/deepke/blob/test_new_deepke/example/ae/standard)**
- The standard module is implemented by common deep learning models, including CNN, RNN, Capsule, GCN, Transformer and the pretrained model.
- Enter the `DeepKE/example/ae/standard` folder.
- The dataset and parameters can be customized in the `data` folder and `conf` folder respectively.
- **Train**: `python run.py`
- **Predict**: `python predict.py`
## Tips
1. Using nearest mirror, like [THU](https://mirrors.tuna.tsinghua.edu.cn/help/anaconda/) in China, will speed up the installation of *Anaconda*.
2. Using nearest mirror, like [aliyun](http://mirrors.aliyun.com/pypi/simple/) in China, will speed up `pip install XXX`.
3. When encountering `ModuleNotFoundError: No module named 'past'`，run `pip install future` .
4. It's slow to install the pretrained language models online. Recommend download pretrained models before use and save them in the `pretrained` folder. Read `README.md` in every task directory to check the specific requirement for saving pretrained models.
## Developers
Zhejiang University: Ningyu Zhang, Liankuan Tao, Haiyang Yu, Xiang Chen, Xin Xu, Xi Tian, Lei Li, Zhoubo Li, Shumin Deng, Yunzhi Yao, Hongbin Ye, Xin Xie, Guozhou Zheng, Huajun Chen
Alibaba DAMO: Chuanqi Tan, Fei Huang