mikemcrae25/black_entity_classifier

🧠 BERT Classifier for Black Article Detection

📝 Model Overview

This repository hosts a fine-tuned BERT model (bert-base-uncased) for classifying sentences as being about Black entities. It is common in social sciences to analyse semantic and topic variation by race. Some racial labels do not require complementary language to disseminate meaning, e.g. "Negro". However, mentions of the word "Black" referencing Black people may be confused with many other meanings. This model identifies if a sentence contains a "Black" entity (i.e. person, group or organisation).

📖 Description

• Model: Fine-tuned bert-base-uncased
• Training Data: 2,000 manually labeled sentences from newspaper articles.
• Inputs: sentence (string)
• Outputs: black_story (0 or 1)

📊 Performance Metrics

• Training Accuracy: 93.5%
• Validation Accuracy: 91.2%
• Precision: 90.8%
• Recall: 92.1%

🚀 Usage Instructions

from transformers import pipeline
classifier = pipeline("text-classification", model="mikemcrae/black-entity-classifier")
result = classifier("Black activists led a peaceful protest downtown.")
print(result)

📊 Example Data Preview (Not actual Newspaper text)

sentence,black_story
"The Black leaders organized a march for civil rights.",1
"The mayor discussed the city's budget for next year.",0
"Black students protested against segregation policies.",1
"Black car for sale.",0

## ⚙️ Reproduction Instructions
```python
from datasets import load_dataset
from transformers import BertForSequenceClassification, Trainer, TrainingArguments

dataset = load_dataset("mikemcrae/black-article-training-data")
model = BertForSequenceClassification.from_pretrained("bert-base-uncased", num_labels=2)

📜 License

MIT License: Free to use with attribution.

MIT License © 2025 Mike McRae
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND.

❤️ Citation

@inproceedings{mcrae2025blackbert,
  title={BERT Classifier for Black Article Detection},
  author={Mike McRae},
  year={2025},
  url={https://huggingface.co/mikemcrae/black-article-classifier}
}