Updated app.py with new sentiment analysis code

README.md

@@ -11,4 +11,99 @@ license: apache-2.0
 short_description: Sentiment Analysis of reviews using Hugging Face Transformer
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# 📝 Multi-Domain Sentiment Analysis
+
+This project is a **Multi-Domain Sentiment Analysis Application** built using [Gradio](https://www.gradio.app/) and [Hugging Face Transformers](https://huggingface.co/transformers/).  
+It allows users to analyze the sentiment of reviews across different domains, such as:
+
+- **General Reviews**
+- **Product Reviews**
+- **Movie Reviews**
+- **Food/Restaurant Reviews**
+
+The app predicts whether the sentiment is **Positive** or **Negative** and shows the confidence score.
+
+---
+
+## 🚀 Features
+- 🌐 **Multi-domain support** (General, Products, Movies, Food/Restaurant).
+- 🤗 **Pretrained models from Hugging Face Hub** for accurate sentiment classification.
+- 🎯 **Confidence Score** included with each prediction.
+- 💡 **User-friendly Gradio Interface** with dropdowns and text inputs.
+- ⚡ **Runs directly in Hugging Face Spaces** — no local setup required.
+
+---
+
+## 🧠 Models Used
+
+Each domain uses a domain-specific model from Hugging Face Hub:
+
+| Domain             | Model Name                                                              | Task                  |
+|--------------------|-------------------------------------------------------------------------|-----------------------|
+| General            | `distilbert-base-uncased-finetuned-sst-2-english`                      | Binary Sentiment      |
+| Product Reviews    | `nlptown/bert-base-multilingual-uncased-sentiment`                     | Star Ratings (mapped) |
+| Movie Reviews      | `textattack/distilbert-base-uncased-imdb`                              | Binary Sentiment      |
+| Food/Restaurant    | `siebert/sentiment-roberta-large-english`                              | Binary Sentiment      |
+
+---
+
+## 📖 How It Works
+
+1. Select a **domain** from the dropdown (e.g., *Movie Reviews*).
+2. Enter a **review** in the text box.
+3. The app returns:
+   - **Domain**
+   - **Predicted Sentiment** (Positive/Negative)
+   - **Confidence Score**
+
+### Example:
+
+Domain: Food/Restaurant
+Sentiment: Negative
+Confidence: 99.95%
+
+
+---
+
+## 🛠️ Installation (Optional: For local testing)
+
+If you want to run this project locally:
+
+```bash
+# Clone the repository
+git clone https://huggingface.co/spaces/<your-username>/multi_domain_reviews_sentiment
+cd multi_domain_reviews_sentiment
+
+# Install dependencies
+pip install -r requirements.txt
+
+# Run the app
+python app.py
+
+
+---
+
+## 📦 Requirements  
+
+The main dependencies are:  
+
+- `transformers`  
+- `torch`  
+- `gradio`  
+
+All dependencies are listed in **`requirements.txt`**.  
+
+---
+
+## 🙌 Acknowledgements  
+
+- 🤗 **Hugging Face** — for models and hosting.  
+- 🎨 **Gradio** — for the simple and elegant UI.  
+- Pretrained models:  
+  - DistilBERT  
+  - NLPTown BERT  
+  - IMDB DistilBERT  
+  - RoBERTa (Siebert)  
+
+
+

app.py

@@ -1,56 +1,46 @@
-# ===============================
-# Sentiment Analysis Web App
-# Using Hugging Face Transformers + Gradio
-# ===============================
+import gradio as gr
+from transformers import pipeline
 
-# Import necessary libraries
-import gradio as gr                     # For building the web UI
-from transformers import pipeline       # For loading pre-trained NLP models
+# Load domain-specific sentiment pipelines
+models={
+    "General":pipeline("sentiment-analysis",model="distilbert-base-uncased-finetuned-sst-2-english"),
+    "Product Reviews":pipeline("sentiment-analysis",model="nlptown/bert-base-multilingual-uncased-sentiment"),
+    "Movie Reviews":pipeline("sentiment-analysis",model="textattack/distilbert-base-uncased-imdb"),
+    "Food/Restaurent":pipeline("sentiment-analysis",model="siebert/sentiment-roberta-large-english"),  
+}
 
-# -------------------------------
-# Step 1: Load Sentiment Model
-# -------------------------------
-# The pipeline function automatically:
-#  - Loads a pre-trained model (by default: distilbert-base-uncased-finetuned-sst-2-english)
-#  - Loads its tokenizer
-#  - Prepares it for inference
-classifier = pipeline("sentiment-analysis")
+# Add label mapping  for IMDB model (Movie Reviews)
+imdb_label_map={
+    "LABEL_0":"Negative",
+    "LABEL_1":"Positive"
+}
 
-# -------------------------------
-# Step 2: Define Prediction Function
-# -------------------------------
-# This function takes a user input (product review text),
-# runs it through the classifier, and formats the output.
-def analyze_sentiment(review):
-    # classifier(review) → returns a list of results (one dict per input string).
-    # Example: [{'label': 'POSITIVE', 'score': 0.9985}]
-    result = classifier(review)[0]  # [0] because we only pass one review at a time.
+# Define sentiment analysis function
+def analyze_sentiment(domain,review):
+    if not review.strip():
+        return "Please enter a review."
+    clf=models[domain]
+    result=clf(review)[0]
     
-    # Format output string with:
-    #  - Sentiment label (POSITIVE/NEGATIVE)
-    #  - Confidence score (rounded to 2 decimal places using .2f)
-    return f"Sentiment: {result['label']} (score: {result['score']:.2f})"
+    # Apply mapping only for Movie Reviews domain as it labesl as LABEL_0 and LABEL_1
+    if domain=="Movie Reviews":
+        result['label']=imdb_label_map[result['label']]
+    
+    return f"Model: {domain}\n\nLabel: {result['label']}\n\nConfidence: {result['score']:.2f}"
 
-# -------------------------------
-# Step 3: Build Gradio Interface
-# -------------------------------
-# Gr.Interface() connects:
-#   - fn → the function to call (our analyze_sentiment function)
-#   - inputs → UI element for user input (Textbox for review text)
-#   - outputs → UI element for showing results (plain text)
-iface = gr.Interface(
+# Create Gradio interface
+iface=gr.Interface(
     fn=analyze_sentiment,
-    inputs=gr.Textbox(
-        lines=3,                          # Multi-line textbox (3 lines height)
-        placeholder="Enter a product review here..."  # Hint text for users
-    ),
-    outputs="text"                        # Output will be shown as plain text
+    inputs=[
+        gr.Dropdown(list(models.keys()),label="Choose Domain",value="General"),
+        gr.Textbox(lines=4,placeholder="Type a review....",label="Review"),  
+    ],
+    outputs=gr.Textbox(label="Prediction"),
+    title="Multi-Domain Sentiment Analysis",
+    description="Pic a domian and analyze sentiment.",
+    allow_flagging="never",
+    
 )
 
-# -------------------------------
-# Step 4: Launch the Web App
-# -------------------------------
-# iface.launch() starts a local web server and
-# provides a shareable link (optional) to access the app.
-iface.launch()
-
+# Launch the interface
+iface.launch()

requirements.txt

@@ -1,3 +1,3 @@
-transformers
-torch
-gradio
+gradio==5.44.1
+transformers==4.44.2
+torch==2.3.0