app.py

import json
import numpy as np
import gradio as gr
from tensorflow import keras
from tensorflow.keras.preprocessing.sequence import pad_sequences

# 1. Load model and tokenizer once at startup
MODEL_PATH = "model/sentiment_model.h5"
WORD_INDEX_PATH = "model/word_index.json"
MAX_LENGTH = 200  # must match training

model = keras.models.load_model(MODEL_PATH)

with open(WORD_INDEX_PATH, "r") as f:
    word_index = json.load(f)

# 2. Preprocessing: raw text -> padded integer sequence
def preprocess(text: str):
    text = text.lower().strip()
    tokens = text.split()

    encoded = [word_index.get(word, 0) + 3 for word in tokens][:MAX_LENGTH]
    padded = pad_sequences([encoded], maxlen=MAX_LENGTH, padding="post")
    return padded

# 3. Prediction logic: padded sequence -> label + confidence
def predict_sentiment(text: str):
    text = text or ""
    if not text.strip():
        return "Please enter some text."

    padded = preprocess(text)
    prob = float(model.predict(padded, verbose=0)[0][0])

    if prob >= 0.5:
        label = "Positive 😀"
        confidence = prob
    else:
        label = "Negative 😞"
        confidence = 1.0 - prob

    return f"{label} (confidence: {confidence:.3f})"

# 4. Build a nice Gradio interface
examples = [
    ["I absolutely loved this movie, it was fantastic!"],
    ["The product is okay, nothing special."],
    ["This was a terrible experience, I am very disappointed."],
    ["The service was quick and the staff were friendly."],
    ["I wouldn't recommend this to anyone."]
]

with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown(
        """
        # 🧠 Sentiment Analysis Demo  
        Type a sentence or review and this neural network will guess whether the sentiment is positive or negative.
        """
    )

    with gr.Row():
        with gr.Column():
            input_box = gr.Textbox(
                lines=4,
                label="Write your text here",
                placeholder="For example: I really enjoyed this movie!"
            )
            submit_btn = gr.Button("Analyze Sentiment", variant="primary")
        with gr.Column():
            label_output = gr.Textbox(
                label="Prediction",
                lines=2
            )

    gr.Examples(
        examples=examples,
        inputs=input_box,
        label="Try one of these examples"
    )

    submit_btn.click(
        fn=predict_sentiment,
        inputs=input_box,
        outputs=label_output
    )

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