Update app.py

app.py

@@ -1,37 +1,45 @@
 import gradio as gr
-from textblob import TextBlob
-import nltk
+from transformers import pipeline
 
+# Load pretrained sentiment analysis pipeline
+sentiment_pipeline = pipeline("sentiment-analysis")
 
-nltk.download('punkt')
 def sentiment_analysis(text: str) -> dict:
     """
-    Analyze the sentiment of the given text.
+    Analyze sentiment using a transformer model, without lexicon.
 
     Args:
-        text (str): The text to analyze
+        text (str): The input text.
 
     Returns:
-        dict: A dictionary containing polarity, subjectivity, and assessment
+        dict: Polarity (mapped from model confidence), assessment, and subjectivity (heuristic).
     """
-    blob = TextBlob(text)
-    sentiment = blob.sentiment
+    result = sentiment_pipeline(text)[0]
     
+    label = result['label']
+    score = result['score']
+
+    polarity = score if label == "POSITIVE" else -score
+    assessment = "positive" if polarity > 0.2 else "negative" if polarity < -0.2 else "neutral"
+    
+    # Heuristic for subjectivity: high confidence implies more subjective
+    subjectivity = round(score, 2)
+
     return {
-        "polarity": round(sentiment.polarity, 2),  # -1 (negative) to 1 (positive)
-        "subjectivity": round(sentiment.subjectivity, 2),  # 0 (objective) to 1 (subjective)
-        "assessment": "positive" if sentiment.polarity > 0 else "negative" if sentiment.polarity < 0 else "neutral"
+        "polarity": round(polarity, 2),
+        "subjectivity": subjectivity,
+        "assessment": assessment
     }
 
-# Create the Gradio interface
+# Gradio Interface
 demo = gr.Interface(
     fn=sentiment_analysis,
-    inputs=gr.Textbox(placeholder="Enter text to analyze..."),
+    inputs=gr.Textbox(placeholder="Enter text to analyze...", lines=4),
     outputs=gr.JSON(),
-    title="Text Sentiment Analysis",
-    description="Analyze the sentiment of text using TextBlob"
+    title="Transformer-Based Sentiment Analysis",
+    description="Analyze sentiment using a pre-trained BERT model (no lexicons or rule-based logic)."
 )
 
-# Launch the interface and MCP server
+# Run on Gradio MCP
 if __name__ == "__main__":
-    demo.launch(mcp_server=True)
+    demo.launch(mcp_server=True)