app.py

import gradio as gr
from transformers import pipeline
import time

print("Loading sentiment model...")
classifier = pipeline("sentiment-analysis", model="distilbert-base-uncased-finetuned-sst-2-english")
print("Model ready!")

def analyze_single(text):
    if not text or not text.strip():
        return "⚠️ Please enter some text.", "", ""

    text = text.strip()
    if len(text) > 512:
        return "⚠️ Text too long. Max 512 characters.", "", ""

    start = time.time()
    result = classifier(text)[0]
    elapsed = round((time.time() - start) * 1000, 1)

    label = result["label"]
    score = round(result["score"] * 100, 2)
    emoji = "😊" if label == "POSITIVE" else "😔"

    sentiment_out = f"{emoji}  {label}"
    confidence_out = f"{score}%"
    time_out = f"{elapsed} ms"

    return sentiment_out, confidence_out, time_out

def analyze_batch(texts_input):
    if not texts_input or not texts_input.strip():
        return "⚠️ Please enter at least one sentence."

    lines = [line.strip() for line in texts_input.strip().split("\n") if line.strip()]

    if len(lines) > 20:
        return "⚠️ Max 20 sentences. Please reduce input."

    results = classifier(lines)

    output_lines = []
    for text, result in zip(lines, results):
        label = result["label"]
        score = round(result["score"] * 100, 1)
        emoji = "😊" if label == "POSITIVE" else "😔"
        short_text = text[:60] + "..." if len(text) > 60 else text
        output_lines.append(f"{emoji} [{label} — {score}%]  {short_text}")

    summary_pos = sum(1 for r in results if r["label"] == "POSITIVE")
    summary_neg = len(results) - summary_pos

    output_lines.append("")
    output_lines.append(f" Summary: {summary_pos} Positive  |  {summary_neg} Negative  |  {len(results)} Total")

    return "\n".join(output_lines)

with gr.Blocks(
    theme=gr.themes.Soft(primary_hue="blue", secondary_hue="indigo"),
    title="Sentiment Analyzer"
) as demo:

    gr.Markdown("""
    #  Sentiment Analyzer
    Detects **Positive** or **Negative** sentiment using DistilBERT.
    Built with HuggingFace Transformers · Model accuracy ~91% on SST-2 benchmark.
    """)

    with gr.Tabs():

        with gr.TabItem("Single Analysis"):
            with gr.Row():
                with gr.Column(scale=2):
                    text_input = gr.Textbox(
                        lines=4,
                        placeholder="Type any sentence here...\ne.g. I love building AI projects!",
                        label="Input Text",
                        max_lines=6
                    )
                    analyze_btn = gr.Button("Analyze Sentiment ", variant="primary", size="lg")

                with gr.Column(scale=1):
                    sentiment_out = gr.Textbox(label="Sentiment", interactive=False)
                    confidence_out = gr.Textbox(label="Confidence Score", interactive=False)
                    time_out = gr.Textbox(label="Response Time", interactive=False)

            gr.Examples(
                examples=[
                    ["I absolutely love building AI projects, it's so rewarding!"],
                    ["This is the worst experience I have ever had."],
                    ["The weather today is absolutely beautiful and I feel great."],
                    ["I am so frustrated and disappointed with this outcome."],
                    ["HuggingFace makes natural language processing incredibly easy!"],
                ],
                inputs=text_input,
                label="Try these examples"
            )

            analyze_btn.click(
                fn=analyze_single,
                inputs=text_input,
                outputs=[sentiment_out, confidence_out, time_out]
            )

        with gr.TabItem("Batch Analysis"):
            gr.Markdown("Enter **one sentence per line** (max 20 sentences)")

            with gr.Row():
                with gr.Column():
                    batch_input = gr.Textbox(
                        lines=8,
                        placeholder="I love this product!\nThis was a terrible experience.\nGreat service, highly recommend.\nI will never buy this again.",
                        label="Input Sentences (one per line)"
                    )
                    batch_btn = gr.Button("Analyze All 🔍", variant="primary", size="lg")

                with gr.Column():
                    batch_output = gr.Textbox(
                        lines=12,
                        label="Results",
                        interactive=False
                    )

            batch_btn.click(
                fn=analyze_batch,
                inputs=batch_input,
                outputs=batch_output
            )

    gr.Markdown("""
    ---
    **Model:** distilbert-base-uncased-finetuned-sst-2-english · **Task:** Binary Sentiment Classification  
    **Developer:** Sawda · BS Computer Engineering Technology · IIU Islamabad
    """)

demo.launch()