metadata
license: cc-by-nc-sa-4.0
明实录与清实录多标签分类推理模型
本模型用于对《明实录》和《清实录》文本进行多标签分类推理。基于Jihuai/bert-ancient-chinese进行任务微调，利用公开语料进行预训练，得到适合实录类型的预训练模型shiluBERT。
中文说明
模型与数据来源
- 训练数据来源：《朝鲜王朝实录》；
- 任务类型：多标签文本分类；
- 训练样本数：约27万。
评估指标
| 指标 | 数值 |
|---|---|
| Sample F1 | 0.7246 |
| Sample Precision | 0.7594 |
| Sample Recall | 0.7321 |
| LRAP | 0.8074 |
| Hamming Loss | 0.0069 |
示例使用方法
- 在线体验 Space： bztxb/shiluInfer
English Version
This model performs multi-label classification inference on texts of VERITABLE RECORDS of the Ming/Qing DYNASTY. It is fine-tuned from Jihuai/bert-ancient-chinese, and further benefits from pretraining on public corpora to obtain a Shilu-oriented pretrained model, shiluBERT.
Model and Data Sources
- Training data source: VERITABLE RECORDS of the JOSEON DYNASTY.
- Task type: multi-label text classification.
- Number of training samples: approximately 0.27 million.
Evaluation Metrics
| Metric | Value |
|---|---|
| Sample F1 | 0.7246 |
| Sample Precision | 0.7594 |
| Sample Recall | 0.7321 |
| LRAP | 0.8074 |
| Hamming Loss | 0.0069 |
Example Usage
- Try the online Space: bztxb/shiluInfer