Erlangshen-Roberta-110M-Sentiment
| Property | Value |
|---|---|
| License | Apache 2.0 |
| Research Paper | View Paper |
| Language | Chinese |
| Training Data | 227,347 samples across 8 datasets |
What is Erlangshen-Roberta-110M-Sentiment?
Erlangshen-Roberta-110M-Sentiment is a specialized Chinese language model based on RoBERTa-wwm-ext-base, fine-tuned specifically for sentiment analysis tasks. Developed by IDEA-CCNL, this model represents a significant advancement in Chinese natural language understanding, particularly in sentiment analysis capabilities.
Implementation Details
The model is built upon the chinese-roberta-wwm-ext-base architecture and has been fine-tuned on an extensive dataset of 227,347 samples from 8 different Chinese sentiment analysis datasets. It demonstrates impressive performance metrics, achieving 97.77% accuracy on ASAP-SENT, 97.31% on ASAP-ASPECT, and 96.61% on ChnSentiCorp benchmarks.
- Built on RoBERTa architecture with 110M parameters
- Implements whole word masking (WWM) technique
- Optimized for Chinese language processing
- Compatible with HuggingFace Transformers library
Core Capabilities
- High-accuracy sentiment analysis for Chinese text
- Robust performance across multiple sentiment analysis benchmarks
- Easy integration with PyTorch-based applications
- Efficient inference with reasonable model size
Frequently Asked Questions
Q: What makes this model unique?
This model stands out due to its specialized fine-tuning on Chinese sentiment analysis tasks and its impressive performance metrics while maintaining a relatively compact size of 110M parameters. It provides an excellent balance between efficiency and accuracy for Chinese sentiment analysis tasks.
Q: What are the recommended use cases?
The model is particularly well-suited for Chinese sentiment analysis applications, including social media monitoring, customer feedback analysis, and opinion mining. It's ideal for scenarios requiring reliable sentiment classification in Chinese text with production-ready performance.
