Add pranalyzer Gradio app

app.py

@@ -0,0 +1,208 @@
+import torch
+import gradio as gr
+from transformers import (
+    pipeline,
+    BartForConditionalGeneration,
+    BartTokenizer
+)
+
+# ── Device ────────────────────────────────────
+DEVICE     = 0 if torch.cuda.is_available() else -1
+DEVICE_STR = "cuda" if torch.cuda.is_available() else "cpu"
+
+# ── Categories & Aspects ──────────────────────
+CATEGORIES = [
+    "Electronics",
+    "Clothing and Fashion",
+    "Books and Literature",
+    "Food and Grocery",
+    "Sports and Outdoors",
+    "Home and Kitchen",
+    "Beauty and Personal Care",
+    "Toys and Games"
+]
+
+ASPECTS = [
+    "price and value",
+    "quality and durability",
+    "delivery and shipping",
+    "customer service",
+    "ease of use",
+    "design and appearance",
+    "performance and speed",
+    "battery life"
+]
+
+# ── Load all models from HuggingFace Hub ──────
+print("Loading sentiment model...")
+sentiment_pipe = pipeline(
+    "text-classification",
+    model  = "Ved2001/pranalyzer",
+    device = DEVICE
+)
+
+print("Loading category classifier...")
+category_pipe = pipeline(
+    "zero-shot-classification",
+    model  = "facebook/bart-large-mnli",
+    device = DEVICE
+)
+
+print("Loading aspect analyzer...")
+aspect_pipe = pipeline(
+    "zero-shot-classification",
+    model  = "cross-encoder/nli-roberta-base",
+    device = DEVICE
+)
+
+print("Loading summarization model...")
+bart_tokenizer = BartTokenizer.from_pretrained(
+    "facebook/bart-large-xsum")
+bart_model = BartForConditionalGeneration\
+    .from_pretrained("facebook/bart-large-xsum")\
+    .to(DEVICE_STR)
+
+print("All models loaded!")
+
+
+# ── Inference functions ───────────────────────
+def analyze_sentiment(text):
+    result = sentiment_pipe(text[:512])[0]
+    return {'label': result['label'], 'score': round(result['score'], 4)}
+
+def classify_category(text):
+    result = category_pipe(
+        text[:512],
+        candidate_labels=CATEGORIES,
+        multi_label=False
+    )
+    return {'category': result['labels'][0], 'score': round(result['scores'][0], 4)}
+
+def analyze_aspects(text):
+    result = aspect_pipe(
+        text[:512],
+        candidate_labels=ASPECTS,
+        multi_label=True
+    )
+    return [
+        (label, round(score, 4))
+        for label, score in zip(result['labels'], result['scores'])
+        if score > 0.3
+    ][:3]
+
+def summarize_review(text):
+    if len(text.split()) < 30:
+        return text
+    inputs = bart_tokenizer(
+        text[:512],
+        return_tensors="pt",
+        truncation=True,
+        max_length=512
+    ).to(DEVICE_STR)
+    summary_ids = bart_model.generate(
+        inputs["input_ids"],
+        max_new_tokens=80,
+        min_length=15,
+        length_penalty=2.0,
+        num_beams=4,
+        early_stopping=True
+    )
+    return bart_tokenizer.decode(
+        summary_ids[0], skip_special_tokens=True)
+
+
+# ── Gradio function ───────────────────────────
+def run_analysis(review_text):
+    if not review_text.strip():
+        return ("Please enter a review!",) * 4
+
+    sentiment = analyze_sentiment(review_text)
+    category  = classify_category(review_text)
+    aspects   = analyze_aspects(review_text)
+    summary   = summarize_review(review_text)
+
+    # Format sentiment
+    emoji = "😊" if sentiment['label'] == "POSITIVE" else "😞"
+    sentiment_out = (
+        f"{emoji} {sentiment['label']}\n"
+        f"Confidence: {sentiment['score']*100:.1f}%"
+    )
+
+    # Format category
+    category_out = (
+        f"📦 {category['category']}\n"
+        f"Confidence: {category['score']*100:.1f}%"
+    )
+
+    # Format aspects
+    aspects_out = "🔍 Aspects Mentioned:\n\n"
+    for aspect, score in aspects:
+        bar   = "█" * int(score * 10)
+        empty = "░" * (10 - int(score * 10))
+        aspects_out += f"• {aspect:<25} {score:.2f}  {bar}{empty}\n"
+    if not aspects:
+        aspects_out += "No strong aspects detected."
+
+    # Format summary
+    summary_out = f"📝 {summary}"
+
+    return sentiment_out, category_out, aspects_out, summary_out
+
+
+# ── Gradio UI ─────────────────────────────────
+with gr.Blocks(title="pranalyzer") as demo:
+
+    gr.Markdown("""
+    # 🛍️ pranalyzer
+    ### Product Review Analyzer
+    Paste any Amazon product review and get instant analysis
+    using 4 NLP models running in parallel.
+    ---
+    """)
+
+    review_input = gr.Textbox(
+        label       = "📝 Paste your product review here",
+        placeholder = "e.g. This laptop is amazing! Battery lasts all day...",
+        lines       = 6
+    )
+
+    analyze_btn = gr.Button(
+        "🔍 Analyze Review",
+        variant = "primary",
+        size    = "lg"
+    )
+
+    gr.Markdown("### 📊 Analysis Results")
+
+    with gr.Row():
+        sentiment_out = gr.Textbox(label="😊 Sentiment", lines=3)
+        category_out  = gr.Textbox(label="📦 Category",  lines=3)
+
+    with gr.Row():
+        aspects_out = gr.Textbox(label="🔍 Aspects", lines=6)
+        summary_out = gr.Textbox(label="📝 Summary", lines=6)
+
+    gr.Examples(
+        examples=[
+            ["This laptop is absolutely incredible! Battery lasts all day, easily 10-12 hours of real work. The display is crisp and bright. Performance is blazing fast. Highly recommended!"],
+            ["Complete waste of money. Stopped working after a week. Customer service was useless and refused a refund. Avoid at all costs."],
+            ["Ordered these running shoes for marathon training. Delivery was super fast. Cushioning is excellent. Only downside is sizing runs small, order a size up."],
+            ["This cookbook is a disappointment. Half the recipes have missing ingredients. Very misleading. Wasted expensive ingredients trying three different recipes."]
+        ],
+        inputs=review_input
+    )
+
+    gr.Markdown("""
+    ---
+    **Models:** `DistilBERT` → Sentiment | `BART-MNLI` → Category | `RoBERTa` → Aspects | `BART-XSUM` → Summary  
+    Built by [Vedant Nagarkar](https://huggingface.co/Ved2001) • 
+    [GitHub](https://github.com/Vedant-Nagarkar/product-review-analyzer)
+    """)
+
+    analyze_btn.click(
+        fn      = run_analysis,
+        inputs  = review_input,
+        outputs = [sentiment_out, category_out, aspects_out, summary_out]
+    )
+
+demo.launch()

requirements.txt

@@ -0,0 +1,10 @@
+transformers==5.0.0
+torch==2.6.0
+gradio==5.50.0
+datasets==4.0.0
+accelerate==1.13.0
+sentencepiece==0.2.1
+huggingface_hub==1.0.0
+evaluate==0.4.6
+scikit-learn==1.6.1
+numpy==1.26.4