app.py

import gradio as gr
import torch
from transformers import DistilBertForSequenceClassification, DistilBertTokenizer

# Load the trained model and tokenizer
model = DistilBertForSequenceClassification.from_pretrained('best_model')
tokenizer = DistilBertTokenizer.from_pretrained('best_model')


# Define the prediction function
def predict_hate_speech(text):
    inputs = tokenizer.encode_plus(
        text,
        add_special_tokens=True,
        max_length=512,
        padding='max_length',
        truncation=True,
        return_tensors='pt'
    )

    input_ids = inputs['input_ids']
    attention_mask = inputs['attention_mask']

    with torch.no_grad():
        outputs = model(input_ids, attention_mask=attention_mask)
        logits = outputs.logits
        probabilities = torch.nn.functional.softmax(logits, dim=-1)
        prediction = torch.argmax(probabilities, dim=1).item()

    labels = {0: 'Neutral', 1: 'Offensive', 2: 'Hateful'}
    predicted_label = labels[prediction]
    confidence_scores = {labels[i]: prob for i, prob in enumerate(probabilities[0].tolist())}

    return predicted_label, confidence_scores

# Define the Gradio interface
interface = gr.Interface(
    fn=predict_hate_speech,
    inputs=gr.Textbox(lines=2, placeholder="Enter text here..."),
    outputs=[
        gr.Textbox(label="Prediction"),
        gr.Label(label="Confidence Scores")
    ],
    title="Hate Speech Detection System using a Deep Active Learning Approach",
    description="Enter a text to predict whether it is Neutral, Offensive, or Hateful.",
    examples=[
        ["I love this product!"],
        ["You are so stupid!"],
        ["I hate this!"]
    ],
    allow_flagging="manual",
    flagging_dir="flagged_data"
)

# Launch the interface
interface.launch()