OneKE
| Property | Value |
|---|---|
| License | CC-BY-NC-SA-4.0 |
| Research Paper | IEPile: Unearthing Large-Scale Schema-Based Information Extraction Corpus |
| Languages | English, Chinese |
| Framework | PyTorch, Transformers |
What is OneKE?
OneKE is a sophisticated bilingual large language model developed jointly by Ant Group and Zhejiang University, specifically designed for knowledge extraction tasks. Built upon the Chinese-Alpaca-2-13B architecture, it offers comprehensive capabilities in named entity recognition (NER), relation extraction (RE), and event extraction (EE) across multiple domains in both Chinese and English.
Implementation Details
The model implements a schema-generalizable information extraction approach, utilizing techniques such as instruction normalization, difficult negative sample collection, and schema-based batched instruction construction. It requires at least 20GB of VRAM for optimal performance and supports both full-parameter and 4-bit quantized inference.
- Supports multiple extraction tasks including NER, RE, EE, and Knowledge Graph Construction
- Implements specialized instruction formats for different extraction tasks
- Provides comprehensive toolchain support for custom schema descriptions
Core Capabilities
- Bilingual knowledge extraction in Chinese and English
- Zero-shot generalization across multiple domains
- Schema-based information extraction with customizable templates
- Support for complex event extraction with multiple arguments
- Integrated knowledge graph construction capabilities
Frequently Asked Questions
Q: What makes this model unique?
OneKE's uniqueness lies in its ability to handle bilingual knowledge extraction tasks with customizable schemas, making it highly adaptable for different domain-specific applications. Its architecture allows for both zero-shot learning and supervised fine-tuning, making it versatile for various use cases.
Q: What are the recommended use cases?
The model is particularly well-suited for medical information extraction, financial report analysis, public sector document processing, and any scenario requiring structured knowledge extraction from unstructured text. It excels in converting complex textual information into structured, machine-readable formats.
