|
|
--- |
|
|
license: mit |
|
|
language: |
|
|
- en |
|
|
- it |
|
|
- es |
|
|
base_model: |
|
|
- microsoft/mdeberta-v3-base |
|
|
pipeline_tag: text-classification |
|
|
metrics: |
|
|
- accuracy |
|
|
library_name: transformers |
|
|
tags: |
|
|
- fact-checking |
|
|
- text-classification |
|
|
--- |
|
|
# GordonAI |
|
|
|
|
|
GordonAI is an AI package designed for sentiment analysis, emotion detection, and fact-checking classification. The models are pre-trained on three languages: **Italian**, **English**, and **Spanish**. |
|
|
|
|
|
## Features |
|
|
This model has been trained specifically for fact-checking tasks. It classifies text into one of four categories: **Disinformation**, **Hoax**, **FakeNews**, or **TrueNews**. |
|
|
|
|
|
Based on the pre-trained mdeberta-v3-base model from Microsoft, it has been fine-tuned on a specialized fact-checking dataset to accurately identify whether a statement is true or false, and to detect misleading or fabricated information. |
|
|
|
|
|
## Usage |
|
|
|
|
|
You can use the `GordonAI` to classify texts helping to identify whether a statement is reliable or misleading. |
|
|
|
|
|
```python |
|
|
from transformers import pipeline |
|
|
|
|
|
# Load the pipeline for text classification |
|
|
classifier = pipeline("text-classification", model="VinMir/GordonAI-fact_checking") |
|
|
|
|
|
# Use the model to classify text |
|
|
result = classifier("The Earth is flat.") |
|
|
print(result) |
|
|
``` |
|
|
|
|
|
## Requirements |
|
|
Python >= 3.9 |
|
|
transformers |
|
|
torch |
|
|
|
|
|
You can install the dependencies using: |
|
|
```bash |
|
|
pip install transformers torch |
|
|
``` |
|
|
## Limitations and bias |
|
|
Please consult the original DeBERTa paper and literature on different NLI datasets for potential biases. |
|
|
|
|
|
## Acknowledgments |
|
|
|
|
|
This package is part of the work for my doctoral thesis. I would like to thank **NeoData** and **Università di Catania** for their valuable contributions to the development of this project. |
