app.py

from transformers import pipeline
import gradio as gr
import lime
import lime.lime_text
import numpy as np
from sklearn.pipeline import make_pipeline

# Load multi-class sentiment analysis model
sentiment_model = pipeline("text-classification", model="cardiffnlp/twitter-roberta-base-sentiment", top_k=None)

# Define possible sentiment classes with a reduced, logical set
label_mapping = {
    "LABEL_0": "negative",
    "LABEL_1": "neutral",
    "LABEL_2": "positive",
    "LABEL_3": "anger",
    "LABEL_4": "chill"
}

# Function to get sentiment prediction
def analyze_sentiment(text):
    results = sentiment_model(text)[0]  # Get predictions
    sorted_results = sorted(results, key=lambda x: x['score'], reverse=True)
    top_label, top_confidence = label_mapping[sorted_results[0]['label']], sorted_results[0]['score']
    
    return f"Sentiment: {top_label} (Confidence: {top_confidence:.2f})"

# Suggest test cases to ensure correct labeling
def get_suggestions():
    return "Try these examples:\n- 'I love this! Best experience ever!' (positive)\n- 'I am so happy today!' (positive)\n- 'It was okay, nothing special.' (neutral)\n- 'I am disappointed with this product.' (negative)\n- 'This is the worst day of my life.' (negative)\n- 'I am furious right now!' (anger)\n- 'I am extremely relaxed and enjoying the moment.' (chill)"

# Explainability function using LIME
def explain_prediction(text):
    explainer = lime.lime_text.LimeTextExplainer(class_names=list(label_mapping.values()))
    
    def predictor(texts):
        predictions = [sentiment_model(text)[0] for text in texts]
        return np.array([[pred[label] if label in pred else 0 for label in label_mapping.values()] for pred in predictions])
    
    exp = explainer.explain_instance(text, predictor, num_features=6)
    return exp.as_list()

# Create Gradio interface
iface = gr.Interface(
    fn=analyze_sentiment, 
    inputs="text", 
    outputs="text",
    title="Multi-Class Sentiment Analysis App",
    description="Enter a sentence to analyze its sentiment across multiple categories (Negative, Neutral, Positive, Anger, Chill).",
    live=True,
    examples=[
        ["I love this! Best experience ever!"],
        ["I am so happy today!"],
        ["It was okay, nothing special."],
        ["I am disappointed with this product."],
        ["This is the worst day of my life."],
        ["The movie was fantastic, I really enjoyed it!"],
        ["I am so angry, I can't believe this happened!"],
        ["I feel completely at peace right now."],
        ["The service was terrible, I wouldn’t recommend this place."],
        ["I feel great today, everything is going well!"],
        ["It’s just another day, nothing special to report."],
        ["This food is awful, I can’t even eat it!"],
        ["The book was so engaging, I couldn’t put it down!"],
        ["I don’t really have an opinion on this matter."],
        ["My day has been okay, not good but not bad either."],
        ["I regret buying this product, it’s a waste of money."],
        ["The customer support was helpful and solved my issue quickly."],
        ["This experience has been quite frustrating, honestly."],
        ["I had fun at the party, it was a great time!"],
        ["There was too much traffic today, it was so annoying."],
        ["I appreciate your help, it really made a difference."],
        ["The test was hard, but I think I did okay."],
        ["I wouldn’t buy this again, it didn’t meet my expectations."],
        ["This new update has improved the app significantly!"],
        ["I’m not sure how I feel about this decision."],
        ["Everything went smoothly today, no issues at all."],
        ["The weather is nice today, not too hot or too cold."],
        ["I had a terrible time at the event, it was poorly organized."],
        ["My experience was neutral, I don’t have strong feelings either way."],
        ["I highly recommend this to everyone, it’s fantastic!"],
        ["This place is so relaxing, I could stay here forever."],
        ["I had a bad day, but I’ll get through it."],
        ["The lecture was informative, I learned a lot."],
        ["It’s neither good nor bad, just okay overall."],
        ["The store was crowded and the staff was rude, not a good experience."],
        ["I’m satisfied with my purchase, it met my expectations."],
        ["This situation is frustrating, I don’t know what to do."],
        ["I’m feeling optimistic about the future!"],
        ["It was a boring day, nothing interesting happened."],
        ["I love spending time with my friends, they make me happy."],
        ["The flight was delayed, but at least I got home safely."],
        ["This dessert is absolutely delicious, I need more!"],
        ["I wish things had gone differently, but it’s okay."],
        ["The staff was unfriendly, I didn’t feel welcome at all."],
        ["I had a productive day, I got a lot of work done."],
        ["This movie was neither exciting nor dull, just in between."],
        ["I’m really grateful for your kindness, it means a lot."],
        ["I have no strong opinion about this, it’s just okay."],
        ["The food was decent, but I’ve had better."],
        ["Everything was perfect, I couldn’t have asked for more!"],
        ["The trip was stressful, nothing went according to plan."],
        ["I’m hopeful that things will get better soon."],
        ["The presentation was well done, I was impressed."],
        ["I feel indifferent about this, it doesn’t affect me much."],
        ["The concert was amazing, I had a blast!"]
    ]
)

iface.launch()