XLM-RoBERTa Japanese NER Model
| Property | Value |
|---|---|
| Parameter Count | 277M |
| License | MIT |
| Model Type | Token Classification |
| Framework | PyTorch with Transformers |
| F1 Score | 0.9864 |
What is xlm-roberta-ner-japanese?
The xlm-roberta-ner-japanese is a specialized Named Entity Recognition (NER) model fine-tuned on Japanese Wikipedia articles. Built upon the xlm-roberta-base architecture, this model excels at identifying and classifying named entities in Japanese text, including persons, organizations, locations, and more.
Implementation Details
The model was fine-tuned using the Stockmark Inc. dataset with carefully selected hyperparameters including a learning rate of 5e-05 and trained over 5 epochs. It utilizes the PyTorch framework and Transformers library, achieving impressive validation metrics with a final F1 score of 0.9864.
- Trained with Adam optimizer (betas=0.9,0.999)
- Batch size of 12 for both training and evaluation
- Linear learning rate scheduler
- Supports 9 distinct entity categories
Core Capabilities
- Person (PER) identification
- Organization detection (ORG, ORG-P, ORG-O)
- Location (LOC) recognition
- Institution/facility (INS) classification
- Product (PRD) and Event (EVT) detection
Frequently Asked Questions
Q: What makes this model unique?
This model stands out for its specialized Japanese NER capabilities, trained on high-quality Wikipedia data with a comprehensive entity classification system. Its high F1 score of 0.9864 demonstrates exceptional accuracy in entity recognition tasks.
Q: What are the recommended use cases?
The model is ideal for Japanese text analysis tasks requiring entity extraction, such as information retrieval, content categorization, and automated text processing. It can be easily implemented using the Transformers pipeline for token classification.
