Update app.py

app.py

@@ -39,29 +39,94 @@ def get_nlp(model_name: str):
         logger.error(f"Error loading model {model_name}: {e}")
         raise gr.Error(f"ไม่สามารถโหลดโมเดล {model_name} ได้: {str(e)}")
 
-# Enhanced label mapping with modern styling for dark blue theme
-LABEL_MAPPINGS = {
-    "LABEL_0": {"code": 0, "name": "question", "emoji": "🤔", "color": "#ef4444", "bg": "rgba(239, 68, 68, 0.2)", "description": "เชิงลบมาก"},
-    "LABEL_1": {"code": 1, "name": "negative", "emoji": "😢", "color": "#f87171", "bg": "rgba(248, 113, 113, 0.2)", "description": "เชิงลบ"},
-    "LABEL_2": {"code": 2, "name": "neutral", "emoji": "😐", "color": "#facc15", "bg": "rgba(250, 204, 21, 0.2)", "description": "เป็นกลาง"},
-    "LABEL_3": {"code": 3, "name": "positive", "emoji": "😊", "color": "#34d399", "bg": "rgba(52, 211, 153, 0.2)", "description": "เชิงบวก"},
-
-    "0": {"code": 0, "name": "negative", "emoji": "😢", "color": "#f87171", "bg": "rgba(248, 113, 113, 0.2)", "description": "เชิงลบ"},
-    "1": {"code": 1, "name": "neutral", "emoji": "😐", "color": "#facc15", "bg": "rgba(250, 204, 21, 0.2)", "description": "เป็นกลาง"},
-    "2": {"code": 2, "name": "positive", "emoji": "😊", "color": "#34d399", "bg": "rgba(52, 211, 153, 0.2)", "description": "เชิงบวก"},
+# Comprehensive label mapping system that covers multiple model formats
+COMPREHENSIVE_LABEL_MAPPINGS = {
+    # Standard format labels (LABEL_X)
+    "LABEL_0": {"sentiment": "negative", "intensity": "normal", "emoji": "😢", "color": "#f87171", "bg": "rgba(248, 113, 113, 0.2)", "description": "เชิงลบ"},
+    "LABEL_1": {"sentiment": "neutral", "intensity": "normal", "emoji": "😐", "color": "#facc15", "bg": "rgba(250, 204, 21, 0.2)", "description": "เป็นกลาง"},
+    "LABEL_2": {"sentiment": "positive", "intensity": "normal", "emoji": "😊", "color": "#34d399", "bg": "rgba(52, 211, 153, 0.2)", "description": "เชิงบวก"},
+    "LABEL_3": {"sentiment": "very_positive", "intensity": "high", "emoji": "🤩", "color": "#22c55e", "bg": "rgba(34, 197, 94, 0.2)", "description": "เชิงบวกมาก"},
+    "LABEL_4": {"sentiment": "question", "intensity": "normal", "emoji": "🤔", "color": "#60a5fa", "bg": "rgba(96, 165, 250, 0.2)", "description": "คำถาม"},
+    
+    # Numeric format labels (0, 1, 2, etc.)
+    "0": {"sentiment": "negative", "intensity": "normal", "emoji": "😢", "color": "#f87171", "bg": "rgba(248, 113, 113, 0.2)", "description": "เชิงลบ"},
+    "1": {"sentiment": "neutral", "intensity": "normal", "emoji": "😐", "color": "#facc15", "bg": "rgba(250, 204, 21, 0.2)", "description": "เป็นกลาง"},
+    "2": {"sentiment": "positive", "intensity": "normal", "emoji": "😊", "color": "#34d399", "bg": "rgba(52, 211, 153, 0.2)", "description": "เชิงบวก"},
+    "3": {"sentiment": "very_positive", "intensity": "high", "emoji": "🤩", "color": "#22c55e", "bg": "rgba(34, 197, 94, 0.2)", "description": "เชิงบวกมาก"},
+    "4": {"sentiment": "question", "intensity": "normal", "emoji": "🤔", "color": "#60a5fa", "bg": "rgba(96, 165, 250, 0.2)", "description": "คำถาม"},
+    
+    # Extended range for models with more classes
+    "5": {"sentiment": "very_negative", "intensity": "high", "emoji": "😡", "color": "#ef4444", "bg": "rgba(239, 68, 68, 0.2)", "description": "เชิงลบมาก"},
+    "LABEL_5": {"sentiment": "very_negative", "intensity": "high", "emoji": "😡", "color": "#ef4444", "bg": "rgba(239, 68, 68, 0.2)", "description": "เชิงลบมาก"},
+    
+    # Text-based labels (common in some models)
+    "NEGATIVE": {"sentiment": "negative", "intensity": "normal", "emoji": "😢", "color": "#f87171", "bg": "rgba(248, 113, 113, 0.2)", "description": "เชิงลบ"},
+    "NEUTRAL": {"sentiment": "neutral", "intensity": "normal", "emoji": "😐", "color": "#facc15", "bg": "rgba(250, 204, 21, 0.2)", "description": "เป็นกลาง"},
+    "POSITIVE": {"sentiment": "positive", "intensity": "normal", "emoji": "😊", "color": "#34d399", "bg": "rgba(52, 211, 153, 0.2)", "description": "เชิงบวก"},
+    "VERY_NEGATIVE": {"sentiment": "very_negative", "intensity": "high", "emoji": "😡", "color": "#ef4444", "bg": "rgba(239, 68, 68, 0.2)", "description": "เชิงลบมาก"},
+    "VERY_POSITIVE": {"sentiment": "very_positive", "intensity": "high", "emoji": "🤩", "color": "#22c55e", "bg": "rgba(34, 197, 94, 0.2)", "description": "เชิงบวกมาก"},
+    "QUESTION": {"sentiment": "question", "intensity": "normal", "emoji": "🤔", "color": "#60a5fa", "bg": "rgba(96, 165, 250, 0.2)", "description": "คำถาม"},
+    
+    # Lowercase variants
+    "negative": {"sentiment": "negative", "intensity": "normal", "emoji": "😢", "color": "#f87171", "bg": "rgba(248, 113, 113, 0.2)", "description": "เชิงลบ"},
+    "neutral": {"sentiment": "neutral", "intensity": "normal", "emoji": "😐", "color": "#facc15", "bg": "rgba(250, 204, 21, 0.2)", "description": "เป็นกลาง"},
+    "positive": {"sentiment": "positive", "intensity": "normal", "emoji": "😊", "color": "#34d399", "bg": "rgba(52, 211, 153, 0.2)", "description": "เชิงบวก"},
+    "very_negative": {"sentiment": "very_negative", "intensity": "high", "emoji": "😡", "color": "#ef4444", "bg": "rgba(239, 68, 68, 0.2)", "description": "เชิงลบมาก"},
+    "very_positive": {"sentiment": "very_positive", "intensity": "high", "emoji": "🤩", "color": "#22c55e", "bg": "rgba(34, 197, 94, 0.2)", "description": "เชิงบวกมาก"},
+    "question": {"sentiment": "question", "intensity": "normal", "emoji": "🤔", "color": "#60a5fa", "bg": "rgba(96, 165, 250, 0.2)", "description": "คำถาม"},
+}
 
+# Sentiment categories for counting and summary
+SENTIMENT_CATEGORIES = {
+    "very_negative": {"name": "เชิงลบมาก", "emoji": "😡", "color": "#ef4444", "order": 1},
+    "negative": {"name": "เชิงลบ", "emoji": "😢", "color": "#f87171", "order": 2},
+    "neutral": {"name": "เป็นกลาง", "emoji": "😐", "color": "#facc15", "order": 3},
+    "positive": {"name": "เชิงบวก", "emoji": "😊", "color": "#34d399", "order": 4},
+    "very_positive": {"name": "เชิงบวกมาก", "emoji": "🤩", "color": "#22c55e", "order": 5},
+    "question": {"name": "คำถาม", "emoji": "🤔", "color": "#60a5fa", "order": 6},
+    "unknown": {"name": "ไม่ทราบ", "emoji": "🔍", "color": "#64748b", "order": 7}
 }
 
 def get_label_info(label: str) -> Dict:
-    """Get label information with fallback for unknown labels"""
-    return LABEL_MAPPINGS.get(label, {
-        "code": -1, 
-        "name": label.lower(), 
+    """
+    Enhanced label information extraction with fallback handling
+    Supports multiple label formats from different models
+    """
+    # Convert to string and clean
+    label_str = str(label).strip()
+    
+    # Try exact match first
+    if label_str in COMPREHENSIVE_LABEL_MAPPINGS:
+        return COMPREHENSIVE_LABEL_MAPPINGS[label_str]
+    
+    # Try case-insensitive match
+    label_upper = label_str.upper()
+    label_lower = label_str.lower()
+    
+    for key, value in COMPREHENSIVE_LABEL_MAPPINGS.items():
+        if key.upper() == label_upper or key.lower() == label_lower:
+            return value
+    
+    # Try pattern matching for complex labels
+    if re.match(r'^LABEL_\d+$', label_str, re.IGNORECASE):
+        # Extract number from LABEL_X format
+        try:
+            num = int(re.findall(r'\d+', label_str)[0])
+            if str(num) in COMPREHENSIVE_LABEL_MAPPINGS:
+                return COMPREHENSIVE_LABEL_MAPPINGS[str(num)]
+        except (IndexError, ValueError):
+            pass
+    
+    # Fallback for unknown labels
+    logger.warning(f"Unknown label: {label}")
+    return {
+        "sentiment": "unknown", 
+        "intensity": "normal",
         "emoji": "🔍", 
         "color": "#64748b", 
         "bg": "rgba(100, 116, 139, 0.2)",
-        "description": "ไม่ทราบ"
-    })
+        "description": f"ไม่ทราบ ({label})"
+    }
 
 def split_sentences(text: str) -> List[str]:
     """Enhanced sentence splitting with better Thai support"""
@@ -82,7 +147,7 @@ def create_confidence_bar(score: float) -> str:
     """
 
 def analyze_text(text: str, model_name: str) -> str:
-    """Enhanced text analysis with modern HTML formatting"""
+    """Enhanced text analysis with comprehensive multi-model support"""
     if not text or not text.strip():
         return """
         <div style="padding: 20px; background: rgba(248, 113, 113, 0.2); border-radius: 12px; border-left: 4px solid #f87171;">
@@ -128,18 +193,11 @@ def analyze_text(text: str, model_name: str) -> str:
     </div>
     """]
     
-    # Updated sentiment_counts to include all 5 sentiment levels plus question
-    sentiment_counts = {
-        "question": 0, 
-        "very negative": 0, 
-        "negative": 0, 
-        "neutral": 0, 
-        "positive": 0, 
-        "very positive": 0, 
-        "other": 0
-    }
+    # Initialize sentiment counting system
+    sentiment_counts = {category: 0 for category in SENTIMENT_CATEGORIES.keys()}
     total_confidence = 0
     sentence_results = []
+    unique_labels_found = set()
     
     # Analyze each sentence
     for i, sentence in enumerate(sentences, 1):
@@ -148,14 +206,17 @@ def analyze_text(text: str, model_name: str) -> str:
             label = result['label']
             score = result['score']
             
+            # Track unique labels for debugging
+            unique_labels_found.add(label)
+            
             label_info = get_label_info(label)
-            label_name = label_info["name"]
+            sentiment_type = label_info["sentiment"]
             
-            # Increment count for the corresponding sentiment name
-            if label_name in sentiment_counts:
-                sentiment_counts[label_name] += 1
+            # Count sentiment
+            if sentiment_type in sentiment_counts:
+                sentiment_counts[sentiment_type] += 1
             else:
-                sentiment_counts["other"] += 1
+                sentiment_counts["unknown"] += 1
             
             total_confidence += score
             
@@ -164,7 +225,8 @@ def analyze_text(text: str, model_name: str) -> str:
                 'sentence': sentence,
                 'label_info': label_info,
                 'score': score,
-                'index': i
+                'index': i,
+                'raw_label': label
             })
             
         except Exception as e:
@@ -207,6 +269,9 @@ def analyze_text(text: str, model_name: str) -> str:
                             <span style="background: {label_info['color']}; color: #f8fafc; padding: 4px 12px; border-radius: 20px; font-size: 12px; font-weight: 600; text-transform: uppercase;">
                                 {label_info['description']}
                             </span>
+                            <span style="color: #64748b; font-size: 12px; background: #1e293b; padding: 2px 8px; border-radius: 10px;">
+                                {result['raw_label']}
+                            </span>
                             <span style="color: #94a3b8; font-size: 14px;">ประโยคที่ {result['index']}</span>
                         </div>
                         <p style="color: #f8fafc; margin: 0 0 12px 0; font-size: 16px; line-height: 1.5;">
@@ -223,55 +288,35 @@ def analyze_text(text: str, model_name: str) -> str:
     total_sentences = len(sentences)
     avg_confidence = total_confidence / total_sentences if total_sentences > 0 else 0
     
-    # Create chart data for summary
+    # Create chart data for summary (only show categories with counts > 0 or important ones)
     chart_items = []
-    # Ensure colors and emojis include all 6 categories (question + 5 sentiments)
-    colors = {
-        "question": "#60a5fa",
-        "very negative": "#ef4444", 
-        "negative": "#f87171", 
-        "neutral": "#facc15", 
-        "positive": "#34d399", 
-        "very positive": "#22c55e",
-        "other": "#64748b"
-    }
-    emojis = {
-        "question": "🤔",
-        "very negative": "😡",
-        "negative": "😢", 
-        "neutral": "😐", 
-        "positive": "😊", 
-        "very positive": "🤩",
-        "other": "🔍"
-    }
+    sorted_categories = sorted(
+        [(k, v, sentiment_counts[k]) for k, v in SENTIMENT_CATEGORIES.items()], 
+        key=lambda x: x[1]["order"]
+    )
     
-    # Iterate through all possible sentiment types to ensure they are displayed, even if count is 0
-    # Or, iterate through sentiment_counts.items() if you only want to display sentiments found in the text
-    for sentiment_name, count in sentiment_counts.items():
-        if sentiment_name == "other" and count == 0: # Don't show 'other' if no unknown labels
-            continue
-        if total_sentences == 0:
-            percentage = 0
-        else:
-            percentage = (count / total_sentences) * 100
+    for sentiment_key, sentiment_info, count in sorted_categories:
+        if count == 0 and sentiment_key == "unknown":
+            continue  # Skip unknown if no unknown sentiments
+            
+        percentage = (count / total_sentences) * 100 if total_sentences > 0 else 0
         
-        # Get description from LABEL_MAPPINGS using the sentiment name
-        # We can find it by iterating through LABEL_MAPPINGS values
-        description = next((info['description'] for info in LABEL_MAPPINGS.values() if info['name'] == sentiment_name), sentiment_name)
-
         chart_items.append(f"""
         <div style="display: flex; align-items: center; gap: 12px; padding: 12px; background: rgba(59, 130, 246, 0.1); border-radius: 8px;">
-            <span style="font-size: 24px;">{emojis.get(sentiment_name, '🔍')}</span>
+            <span style="font-size: 24px;">{sentiment_info['emoji']}</span>
             <div style="flex: 1;">
-                <div style="font-weight: 600; color: #f8fafc; text-transform: capitalize;">{description}</div>
+                <div style="font-weight: 600; color: #f8fafc;">{sentiment_info['name']}</div>
                 <div style="color: #94a3b8; font-size: 14px;">{count} ประโยค ({percentage:.1f}%)</div>
             </div>
             <div style="width: 60px; height: 6px; background: #334155; border-radius: 3px; overflow: hidden;">
-                <div style="width: {percentage}%; height: 100%; background: {colors.get(sentiment_name, '#64748b')}; transition: all 0.3s ease;"></div>
+                <div style="width: {percentage}%; height: 100%; background: {sentiment_info['color']}; transition: all 0.3s ease;"></div>
             </div>
         </div>
         """)
     
+    # Debug information showing found labels
+    debug_labels = ", ".join(sorted(unique_labels_found))
+    
     html_parts.append(f"""
     <div style="padding: 24px; background: linear-gradient(135deg, #1e293b 0%, #0f172a 100%);">
         <h3 style="color: #f8fafc; margin: 0 0 20px 0; font-size: 20px; font-weight: 700; display: flex; align-items: center; gap: 8px;">
@@ -290,9 +335,14 @@ def analyze_text(text: str, model_name: str) -> str:
             </div>
         </div>
         
-        <div style="display: grid; gap: 8px;">
+        <div style="display: grid; gap: 8px; margin-bottom: 16px;">
             {"".join(chart_items)}
         </div>
+        
+        <div style="background: #1e293b; padding: 12px; border-radius: 8px; border-left: 4px solid #60a5fa;">
+            <div style="color: #60a5fa; font-size: 12px; font-weight: 600; margin-bottom: 4px;">🔍 Labels ที่พบในโมเดลนี้:</div>
+            <div style="color: #94a3b8; font-size: 12px; font-family: monospace;">{debug_labels}</div>
+        </div>
     </div>
     """)
     
@@ -433,8 +483,8 @@ with gr.Blocks(css=CUSTOM_CSS, theme=gr.themes.Base()) as demo:
         with gr.Row():
             gr.HTML("""
             <div class='main-uxui-header'>
-                <h1>Thai Sentiment Analysis (SpaceThai-e5)</h1>
-                <p>วิเคราะห์ความรู้สึกภาษาไทย/อังกฤษ รองรับหลายโมเดล | Modern UX/UI</p>
+                <h1>Thai Sentiment Analysis (Multi-Model)</h1>
+                <p>วิเคราะห์ความรู้สึกภาษาไทย/อังกฤษ รองรับหลายโมเดล | Enhanced Multi-Label Support</p>
             </div>
             """)
         with gr.Row():
@@ -462,8 +512,10 @@ with gr.Blocks(css=CUSTOM_CSS, theme=gr.themes.Base()) as demo:
                 ["เศร้ามากเลยวันนี้ งานเยอะเกินไป"],
                 ["อาหารอร่อยดี แต่บริการช้ามาก"],
                 ["คุณคิดอย่างไรกับเศรษฐกิจไทย?"],
-                ["I love this product! It's amazing."],
-                ["This is the worst experience I've ever had."]
+                ["I love this product! It's amazing and perfect!"],
+                ["This is the worst experience I've ever had. Absolutely terrible!"],
+                ["The weather is okay today, nothing special."],
+                ["What do you think about this new technology?"]
             ],
                 inputs=input_box,
                 label="ตัวอย่างข้อความ",
@@ -472,15 +524,25 @@ with gr.Blocks(css=CUSTOM_CSS, theme=gr.themes.Base()) as demo:
             gr.HTML("""
             <div class='main-uxui-legend'>
                 <div class='main-uxui-section-title'>
-                    <span>🗂️</span> คำอธิบายผลลัพธ์
+                    <span>🗂️</span> คำอธิบายผลลัพธ์ (รองรับ Label หลายรูปแบบ)
                 </div>
                 <div class='legend-row'>
-                    <div class='legend-item'><strong>🤩 เชิงบวกมาก</strong><br><small>Very Positive</small></div>
-                    <div class='legend-item'><strong>😊 เชิงบวก</strong><br><small>Positive</small></div>
-                    <div class='legend-item'><strong>😐 เป็นกลาง</strong><br><small>Neutral</small></div>
-                    <div class='legend-item'><strong>😢 เชิงลบ</strong><br><small>Negative</small></div>
-                    <div class='legend-item'><strong>😡 เชิงลบมาก</strong><br><small>Very Negative</small></div>
-                    <div class='legend-item'><strong>🤔 คำถาม</strong><br><small>Question</small></div>
+                    <div class='legend-item'><strong>😡 เชิงลบมาก</strong><br><small>Very Negative<br>Labels: 5, LABEL_5, VERY_NEGATIVE</small></div>
+                    <div class='legend-item'><strong>😢 เชิงลบ</strong><br><small>Negative<br>Labels: 0, LABEL_0, NEGATIVE</small></div>
+                    <div class='legend-item'><strong>😐 เป็นกลาง</strong><br><small>Neutral<br>Labels: 1, LABEL_1, NEUTRAL</small></div>
+                    <div class='legend-item'><strong>😊 เชิงบวก</strong><br><small>Positive<br>Labels: 2, LABEL_2, POSITIVE</small></div>
+                    <div class='legend-item'><strong>🤩 เชิงบวกมาก</strong><br><small>Very Positive<br>Labels: 3, LABEL_3, VERY_POSITIVE</small></div>
+                    <div class='legend-item'><strong>🤔 คำถาม</strong><br><small>Question<br>Labels: 4, LABEL_4, QUESTION</small></div>
+                </div>
+                <div style="margin-top: 16px; padding: 12px; background: rgba(96, 165, 250, 0.1); border-radius: 8px; border-left: 4px solid #60a5fa;">
+                    <div style="color: #60a5fa; font-weight: 600; margin-bottom: 8px;">✨ คุณสมบัติใหม่:</div>
+                    <ul style="color: #94a3b8; font-size: 14px; margin: 0; padding-left: 20px;">
+                        <li>รองรับ Label หลายรูปแบบ (LABEL_X, ตัวเลข, ข้อความ)</li>
+                        <li>แสดง Label ดิบที่โมเดลส่งออกมา</li>
+                        <li>ระบบ Fallback สำหรับ Label ที่ไม่รู้จัก</li>
+                        <li>Debug information แสดง Label ที่พบ</li>
+                        <li>การนับและสรุปผลที่แม่นยำขึ้น</li>
+                    </ul>
                 </div>
             </div>
             """)