app.py

import gradio as gr
import pickle
import os
import numpy as np

# Define paths to the pickle files
BASE_DIR = os.path.dirname(os.path.abspath(__file__))
PKL_DIR = os.path.join(BASE_DIR, 'pkl_files')

SENTIMENT_MODEL_PATH = os.path.join(PKL_DIR, 'sentiment_model.pkl')
SENTIMENT_VECTORIZER_PATH = os.path.join(PKL_DIR, 'sentiment_vectorizer.pkl')
GENRE_MODEL_PATH = os.path.join(PKL_DIR, 'genre_model.pkl')
GENRE_VECTORIZER_PATH = os.path.join(PKL_DIR, 'genre_vectorizer.pkl')

# Load models and vectorizers
def load_pickle(path):
    with open(path, 'rb') as f:
        return pickle.load(f)

print("Loading models...")
try:
    sentiment_model = load_pickle(SENTIMENT_MODEL_PATH)
    sentiment_vectorizer = load_pickle(SENTIMENT_VECTORIZER_PATH)
    genre_model = load_pickle(GENRE_MODEL_PATH)
    genre_vectorizer = load_pickle(GENRE_VECTORIZER_PATH)
    print("Models loaded successfully.")
except Exception as e:
    print(f"Error loading models: {e}")
    raise e

def predict_review(review_text):
    if not review_text:
        return "Please enter a review.", "Please enter a review."
    
    # Sentiment Prediction
    try:
        # Transform text using the sentiment vectorizer
        # Note: transform expects an iterable, so we wrap review_text in a list
        sentiment_features = sentiment_vectorizer.transform([review_text])
        sentiment_prediction = sentiment_model.predict(sentiment_features)[0]
    except Exception as e:
        sentiment_prediction = f"Error in sentiment prediction: {str(e)}"

    # Genre Prediction
    try:
        # Transform text using the genre vectorizer
        genre_features = genre_vectorizer.transform([review_text])
        
        # Use decision_function to get confidence scores
        if hasattr(genre_model, "decision_function"):
            decision_scores = genre_model.decision_function(genre_features)[0]
            
            # Apply Softmax to convert scores to probabilities
            exp_scores = np.exp(decision_scores)
            probabilities = exp_scores / np.sum(exp_scores)
            
            # Map classes to probabilities
            genre_prediction = {
                label: float(prob) 
                for label, prob in zip(genre_model.classes_, probabilities)
            }
        else:
            # Fallback if decision_function is not available
            pred = genre_model.predict(genre_features)[0]
            genre_prediction = {pred: 1.0}
            
    except Exception as e:
        genre_prediction = {"Error": str(e)}

    return sentiment_prediction, genre_prediction

# Create Gradio Interface
iface = gr.Interface(
    fn=predict_review,
    inputs=gr.Textbox(lines=5, placeholder="Enter movie review here...", label="Movie Review"),
    outputs=[
        gr.Textbox(label="Predicted Sentiment"),
        gr.Label(label="Predicted Genre", num_top_classes=3)
    ],
    title="Movie Review Sentiment & Genre Classifier",
    description="Enter a movie review to predict its sentiment and the movie genre based on the text."
)

if __name__ == "__main__":
    iface.launch()