DeepKE is a knowledge extraction toolkit supporting cnSchema, standard supervised, low-resource and document-level scenarios for entity, relation and attribution extraction. It allows developers and researchers to customize datasets and models to extract information from unstructured texts.
DeepKE supports low-resource setting with only a few labelled (e.g., 16/32 shot) instances for widespread information extraction tasks.
Since relations are expressed over multiple sentences in real-world applications, DeepKE supports document-level relation extraction.
DeepKE supports multimodal entity and relation extraction tasks, which can enhance the extraction performance through visual cues.
We provide an online system to extract knowledge from the text with friendly interactive interfaces and fast reaction speed. Click here to try it !!
We present a new open-source and extensible knowledge extraction toolkit, called DeepKE, supporting standard fully supervised, low-resource few-shot and document-level scenarios. DeepKE implements various information extraction tasks, including named entity recognition, relation extraction and attribute extraction. With a unified framework, DeepKE allows developers and researchers to customize datasets and models to extract information from unstructured texts according to their requirements. Specifically, DeepKE not only provides various functional modules and model implementation for different functions and scenarios but also organizes all components by consistent frameworks to maintain sufficient modularity and extensibility. Moreover, DeepKE has quipped with comprehensive documents as well as Google Colab tutorials for beginners. Users can install DeepKE via 'install deepke'. We will provide maintenance to meet new requests, add new tasks, and fix bugs in the future.
DeepKE provides various functional modules and reganizes all components by consistent frameworks. DeepKE provides off-the-shelf extraction models with Chinese pre-trained language models based cnSchema.
The training & evaluation codes and model implementation are separated for easy usage.
An off-the-shelf automatic hyperparameter tuning component is available.
Ningyu Zhang, Liankuan Tao, Xin Xu, Haiyang Yu, Hongbin Ye, Shuofei Qiao, Xin Xie,Xinrong Li, Xiang Chen, Zhoubo Li, Lei Li, Xiaozhuan Liang, YunzhiYao, Shumin Deng, Peng Wang, Wen Zhang, Guozhou Zheng, Huajun Chen
Qiang Chen, Feiyu Xiong
Zhenru Zhang, Chuanqi Tan, Fei Huang