Start

Model Framework

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, few-shot, document-level and multimodal submodules in the relation 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.

Dataset

We use the following datasets in our experiments:

Task	Settings	Corpus	Language	Model
Name Entity Recognition	Standard	CoNLL-2003	English	BERT
		People’s Daily	Chinese
	Few-shot	CoNLL-2003	English	LightNER
		MIT Movie
		MIT Restaurant
		ATIS
	Multimodal	Twitter15	English	IFAformer
		Twitter17
Relation Extraction	Standard	DuIE	Chinese	CNN
				RNN
				Capsule
				GCN
				Transformer
				BERT
	Few-shot	SEMEVAL(8-shot)	English	KnowPrompt
		SEMEVAL(16-shot)
		SEMEVAL(32-shot)
		SEMEVAL(Full)
	Document	DocRED	English	DocuNet
		CDR
		GDA
	Multimodal	MNRE	English	IFAformer
Attribute Extraction	Standard	Triplet Extraction Dataset	Chinese	CNN
				RNN
				Capsule
				GCN
				Transformer
				BERT

Get Start

If you want to use our code , you can do as follow:

git clone https://github.com/zjunlp/DeepKE.git
cd DeepKE