Randeng-Pegasus-523M-Summary-Chinese
| Property | Value |
|---|---|
| Model Size | 523M parameters |
| Task | Text Summarization |
| Language | Chinese |
| Research Paper | PEGASUS Paper |
| Framework | PyTorch |
What is Randeng-Pegasus-523M-Summary-Chinese?
Randeng-Pegasus-523M-Summary-Chinese is a specialized Chinese language model designed for text summarization tasks. It's based on the PEGASUS architecture and has been fine-tuned on seven diverse Chinese datasets totaling approximately 4M samples, including education, news, social media, and general content sources.
Implementation Details
The model is built upon the PEGASUS architecture and has been specifically optimized for Chinese text summarization. It achieves impressive performance metrics, including a ROUGE-1 score of 48.00 on the LCSTS benchmark, demonstrating its effectiveness in generating high-quality Chinese summaries.
- Architecture: PEGASUS-large adapted for Chinese language
- Training Data: Fine-tuned on 7 datasets (education, new2016zh, nlpcc, shence, sohu, thucnews, weibo)
- Integration: Easily implementable using HuggingFace's transformers library
Core Capabilities
- Automated Chinese text summarization
- High-quality abstract generation
- Efficient processing of long-form Chinese content
- Strong performance on standard benchmarks
Frequently Asked Questions
Q: What makes this model unique?
This model stands out for its specialized focus on Chinese text summarization, combining the powerful PEGASUS architecture with extensive fine-tuning on diverse Chinese datasets. Its impressive ROUGE scores demonstrate its effectiveness in generating high-quality Chinese summaries.
Q: What are the recommended use cases?
The model is ideal for applications requiring Chinese text summarization, including news digestion, content curation, document summarization, and automated abstract generation. It's particularly well-suited for processing longer Chinese texts where concise summaries are needed.
