Update app.py

app.py

@@ -25,7 +25,7 @@ humanized_map = {
     "LABEL_5": "😲 বিস্ময় (Surprise)",
     "LABEL_6": "😐 নিরপেক্ষ (Neutral)",
 
-    # Just in case model returns plain text instead of label_*
+    # Extra fallback for plain text labels
     "Anger": "😠 রাগ (Anger)",
     "Sadness": "😢 দুঃখ (Sadness)",
     "Joy": "😊 আনন্দিত (Joy)",
@@ -35,6 +35,23 @@ humanized_map = {
     "Neutral": "😐 নিরপেক্ষ (Neutral)"
 }
 
+# Extra keyword-based fallback mapping
+keyword_map = {
+    "happy": "😊 আনন্দিত (Joy)",
+    "খুশি": "😊 আনন্দিত (Joy)",
+    "sad": "😢 দুঃখ (Sadness)",
+    "দুঃখ": "😢 দুঃখ (Sadness)",
+    "love": "❤️ ভালোবাসা (Love)",
+    "ভালবাসা": "❤️ ভালোবাসা (Love)",
+    "ভালোবাসা": "❤️ ভালোবাসা (Love)",
+    "fear": "😨 ভয় (Fear)",
+    "ভয়": "😨 ভয় (Fear)",
+    "angry": "😠 রাগ (Anger)",
+    "রাগ": "😠 রাগ (Anger)",
+    "surprise": "😲 বিস্ময় (Surprise)",
+    "বিস্ময়": "😲 বিস্ময় (Surprise)"
+}
+
 # Emotion detect function
 def detect_emotion(text):
     if not text.strip():
@@ -44,10 +61,15 @@ def detect_emotion(text):
         label = result["label"]
         score = round(result["score"] * 100, 2)
         emotion = humanized_map.get(label)
-        if emotion:
-            return f"{emotion} (score: {score}%)"
-        else:
+
+        if not emotion:
+            # Keyword fallback
+            for word in keyword_map:
+                if word in text.lower():
+                    return f"{keyword_map[word]} (keyword match)"
             return f"🤔 অজানা (Unknown) — মডেল লেবেল: {label} (score: {score}%)"
+
+        return f"{emotion} (score: {score}%)"
     except Exception as e:
         return f"❌ সমস্যা হয়েছে: {str(e)}"