metadata
language:
- en
license: apache-2.0
tags:
- deberta
- text-classification
- microaggression
- detection
- bias
pipeline_tag: text-classification
widget:
- text: You speak good English for someone from there.
- text: Where are you really from?
- text: You're so articulate.
datasets:
- custom
metrics:
- accuracy
- f1
model-index:
- name: CI_MA_Detect
results:
- task:
type: text-classification
name: Microaggression Detection
metrics:
- type: accuracy
value: 0.85
name: Accuracy
CI_MA_Detect - Microaggression Detection Model
This model detects microaggressions in text using a fine-tuned DeBERTa architecture.
Model Description
- Model type: DeBERTa for sequence classification
- Task: Binary text classification (microaggression detection)
- Labels:
- LABEL_0: Not a microaggression
- LABEL_1: Microaggression detected
Usage
from transformers import DebertaTokenizer, DebertaForSequenceClassification
import torch
tokenizer = DebertaTokenizer.from_pretrained("jokugeorgin/CI_MA_Detect")
model = DebertaForSequenceClassification.from_pretrained("jokugeorgin/CI_MA_Detect")
text = "You speak good English for someone from there."
inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=128)
outputs = model(**inputs)
prediction = torch.argmax(outputs.logits, dim=1)
API Usage
curl https://api-inference.huggingface.co/models/jokugeorgin/CI_MA_Detect \
-H "Authorization: Bearer YOUR_HF_TOKEN" \
-H "Content-Type: application/json" \
-d '{"inputs": "You speak good English for someone from there."}'
Training Data
Custom dataset of microaggression examples and neutral text.
Limitations
- Works best with English text
- May require context for ambiguous statements
- Performance varies with text length and complexity