| license: mit | |
| language: | |
| - en | |
| library_name: transformers | |
| pipeline_tag: text-classification | |
| tags: | |
| - ethics | |
| - research | |
| - deberta | |
| - classification | |
| datasets: | |
| - custom | |
| metrics: | |
| - accuracy | |
| - f1 | |
| base_model: microsoft/deberta-v3-base | |
| model-index: | |
| - name: ethics-review-deberta | |
| results: [] | |
| # Ethics Review DeBERTa Model | |
| A fine-tuned DeBERTa-v3-base model for classifying research ethics guideline compliance. | |
| ## Model Description | |
| This model classifies text segments from research proposals to determine if they adequately address specific ethics review guidelines. | |
| ## Labels | |
| - **LABEL_0 / ADDRESSED**: The text adequately addresses the ethics guideline | |
| - **LABEL_1 / NEEDS_REVISION**: The text needs revision or doesn't address the guideline | |
| ## Usage | |
| ```python | |
| from transformers import pipeline | |
| classifier = pipeline("text-classification", model="JohnLicode/ethics-review-deberta") | |
| # Example | |
| text = "Guideline 1.1 Objectives: The general objective is to develop an AI ethics review system." | |
| result = classifier(text) | |
| print(result) | |
| # [{'label': 'LABEL_0', 'score': 0.95}] # ADDRESSED | |
| ``` | |
| ## Training | |
| - Base model: `microsoft/deberta-v3-base` | |
| - Task: Binary text classification | |
| - Training data: Custom ethics review dataset with 30 guideline categories | |
| ## Intended Use | |
| This model is designed to assist ethics review committees by providing preliminary assessments of research proposals against institutional guidelines. | |