DeepKE

DeepKE is a deep learning toolkit that extracts entities, relations, and attributes from unstructured text to build knowledge graphs.

DeepKE is an open-source, extensible toolkit designed for knowledge extraction tasks including named entity recognition, relation extraction, and attribute extraction. The framework enables developers and researchers to customize models and datasets to pull information from unstructured text. It supports diverse scenarios such as standard supervised learning, low-resource settings requiring few-shot instances, and document-level relation extraction where relationships span multiple sentences. Additionally, the toolkit supports multi-modal extraction to incorporate visual cues alongside text. The system is organized with modularity in mind, separating training and evaluation codes, and offers automatic hyperparameter tuning.

DeepKE at a glance

Best for: Researcher, Software engineer
Runs on: open source
Pricing: Free
Released: August 2018
Category: AI research tool
Website: deepke.zjukg.cn

Key features

Supports named entity recognition, relation, and attribute extraction
Enables low-resource, few-shot extraction settings
Handles document-level relation extraction
Supports multi-modal entity and relation extraction
Provides automatic hyperparameter tuning
Integrates cnSchema with pre-trained Chinese language models
Organizes components with modular framework architecture

Use it for

Building knowledge bases from unstructured text
Performing named entity recognition on text data
Extracting relations across document-level spans
Developing models for low-resource information extraction

Alternatives to DeepKE

GLiNER — Enables zero-shot entity extraction without labeled training data on consumer CPUs.
AI Knowledge Graph — Extracts knowledge from documents and builds interactive, traceable graphs.
Hyper-Extract — Extracts complex knowledge graphs and hypergraphs from documents with one command.
LangExtract — It maps every extraction to its exact location in the source text.
GraphRAG — Enhances LLM reasoning over private data using knowledge graph structures.
DeepSearcher — Performs evaluation and reasoning based on private enterprise data using various models.
DocETL — Extracts structured data and links knowledge using large language models.
WeKnora — Combines RAG, multi-step agents, and automatic Wiki generation in one platform.
deepdoctection — Orchestrates layout analysis, OCR, and model fine-tuning in PyTorch pipelines.
LightRAG — It uses a dual-layer architecture to reduce LLM calls and retrieval latency.
nerfguard-extract — Delivers frontier quality extraction for lower cost and faster speed.
DocTree — Generates source-grounded learning articles adapted to your expertise and time

All DeepKE alternatives