Update app.py

app.py

@@ -3,25 +3,24 @@ from indic_transliteration import sanscript
 from indic_transliteration.sanscript import transliterate
 from transformers import pipeline
 
-# 1. Load public models
-# Public Hindi sentiment & emotion models (authentication-free)
+# 1. Emotion और Sentiment models (public, authentication-free)
 emotion_model = pipeline(
     "text-classification",
-    model="l3cube-pune/hindi-emotion",  # Public Hindi emotion model
+    model="j-hartmann/emotion-english-distilroberta-base",
     return_all_scores=True
 )
 
 sentiment_model = pipeline(
     "text-classification",
-    model="nlptown/bert-base-multilingual-uncased-sentiment",  # Multilingual sentiment model
+    model="nlptown/bert-base-multilingual-uncased-sentiment",
     return_all_scores=True
 )
 
-# 2. Hinglish -> Hindi
+# 2. Hinglish -> Hindi transliteration
 def hinglish_to_hindi(text: str) -> str:
     return transliterate(text, sanscript.ITRANS, sanscript.DEVANAGARI)
 
-# 3. Normalization
+# 3. Basic Hindi normalization
 def normalize_hindi(text: str) -> str:
     corrections = {
         "उदस्": "उदास",
@@ -34,47 +33,49 @@ def normalize_hindi(text: str) -> str:
         text = text.replace(wrong, right)
     return text
 
-# 4. Emoji mapping
+# 4. Emoji mapping for quick visual
 EMOJI_MAP = {
-    "anger":"😡","disgust":"🤢","fear":"😱","joy":"😄","neutral":"😐","sadness":"😢","surprise":"😲"
+    "anger":"😡","disgust":"🤢","fear":"😱","joy":"😄",
+    "neutral":"😐","sadness":"😢","surprise":"😲"
 }
 
 SENTIMENT_MAP = {
-    "1":"😞 Negative",
-    "2":"😐 Slightly Negative",
-    "3":"😐 Neutral",
-    "4":"😃 Positive",
-    "5":"😃 Very Positive"
+    "1 star":"😞 Negative",
+    "2 stars":"😟 Negative",
+    "3 stars":"😐 Neutral",
+    "4 stars":"🙂 Positive",
+    "5 stars":"😃 Positive"
 }
 
-# 5. Full pipeline
-def analyze_text_visual(hinglish_text: str):
+# 5. Complete pipeline
+def analyze_text(hinglish_text: str):
+    # Transliterate + normalize
     hindi_text = normalize_hindi(hinglish_to_hindi(hinglish_text))
     
-    # Emotion analysis
-    emotion_scores = emotion_model(hindi_text)
-    top_emotion = max(emotion_scores[0], key=lambda x: x['score'])
+    # Emotion prediction
+    emotions = emotion_model(hindi_text)[0]
+    top_emotion = max(emotions, key=lambda x: x['score'])
+    emotion_label = top_emotion['label']
+    emotion_score = top_emotion['score']
+    emoji = EMOJI_MAP.get(emotion_label.lower(), "❓")
     
-    # Sentiment analysis
-    sentiment_scores = sentiment_model(hindi_text)
-    top_sentiment = max(sentiment_scores[0], key=lambda x: x['score'])
+    # Sentiment prediction
+    sentiments = sentiment_model(hindi_text)[0]
+    top_sentiment = max(sentiments, key=lambda x: x['score'])
+    sentiment_label = top_sentiment['label']
+    sentiment_score = top_sentiment['score']
+    sentiment_display = SENTIMENT_MAP.get(sentiment_label, sentiment_label)
     
-    # Prepare user-friendly output
-    output = {
-        "Hinglish Input": hinglish_text,
-        "Hindi Text": hindi_text,
-        "Top Emotion": f"{EMOJI_MAP.get(top_emotion['label'], '')} {top_emotion['label']} ({top_emotion['score']:.2f})",
-        "Top Sentiment": f"{SENTIMENT_MAP.get(str(int(top_sentiment['label'][-1])), '')} ({top_sentiment['score']:.2f})"
-    }
-    return output
+    # Return readable summary
+    return f"हिसाब से भावनाएँ: {emotion_label} {emoji} ({emotion_score:.2f})\nसेंटिमेंट: {sentiment_display} ({sentiment_score:.2f})"
 
 # 6. Gradio interface
 iface = gr.Interface(
-    fn=analyze_text_visual,
+    fn=analyze_text,
     inputs="text",
-    outputs="json",
-    title="Hinglish → Hindi Emotion & Sentiment Analysis",
-    description="Type Hinglish text, see Hindi transliteration, top emotion and sentiment with emojis."
+    outputs="text",
+    title="Hinglish → Hindi Emotion & Sentiment Detector",
+    description="Hinglish या Hindi text डालें, परिणाम emoji और readable format में मिलेगा।"
 )
 
 iface.launch()