peekayitachi/BiasCheck-RoBERTa

BiasCheck-RoBERTa

Model Description

The BiasCheck-RoBERTa model is a political bias detection model based on the RoBERTa architecture. It classifies news articles into three political bias categories: Left, Center, and Right. This model was trained on a curated dataset of articles from allsides available on kaggle, and it utilizes the RoBERTa-base model as the base architecture for text classification. The model provides a reliable way to identify political bias in news articles, helping users to assess the bias of the content they consume.

Base Model

The BiasCheck-RoBERTa model is based on the RoBERTa-base architecture, a robust transformer-based model that has been pre-trained on vast amounts of text data.

License

This model is licensed under the MIT License.

Training Data

The model was trained on the following datasets:

• —AllSides Ratings of Bias in Electronic Media
• —Article Bias Prediction Dataset

Metrics

The model was evaluated using several performance metrics. Below are the key metrics:

• —Accuracy: 0.913
• —Precision: 0.914
• —Recall: 0.913
• —F1-Score: 0.913
• —Log Loss: 0.233
• —AUC-ROC: 0.986

Carbon Emission

Experiments were conducted using a private infrastructure, which has a carbon efficiency of 0.432 kgCO$_2$eq/kWh. A cumulative of 24 hours of computation was performed on hardware of type RTX 3080 (TDP of 320W).

Total emissions are estimated to be 3.32 kgCO2eq of which 0 percents were directly offset.

3.32 kgCO2eq is equivalent to

• —13.4 Km driven by an average ICE car
• —1.66 Kgs of coal burned
• —0.06 Tree seedlings sequesting carbon for 10 years

Installation

To use this model, you will need to install the following dependencies:

make a requirements.txt file and add the following dependencies <br />

torch>=2.0 <br /> transformers>=4.35 <br /> datasets <br /> scikit-learn <br /> matplotlib <br /> numpy <br /> pandas <br /> nltk<br /> spacy<br /> pydantic<br /> fastapi<br /> uvicorn<br /> bert-score<br /> contractions<br /> torch<br /> transformers<br /> datasets<br /> scikit-learn<br /> pandas<br /> numpy<br /> matplotlib<br /> seaborn<br /> bert_score<br />

pip install -r requirements.txt