Update app.py

app.py

@@ -1,72 +1,77 @@
-import gradio as gr
-from transformers import pipeline
-
-# 1. LOAD MODEL NLP UNTUK SENTIMEN
-print("🚀 Sedang memuat model NLP (RoBERTa)...")
-# top_k=None memastikan model mengembalikan skor probabilitas untuk ketiga kelas (Positif, Netral, Negatif)
-sentiment_pipeline = pipeline("text-classification", model="cardiffnlp/twitter-roberta-base-sentiment-latest", top_k=None)
-
-def analyze_sentiment(text):
-    if not text.strip():
-        return "⚠️ Teks kosong", {"😐 Netral": 1.0}
-    
-    try:
-        # AI memproses kalimat
-        results = sentiment_pipeline(text)[0]
-        
-        # Mapping label bahasa Inggris ke visualisasi UI bahasa Indonesia
-        label_mapping = {
-            "positive": "😊 Positif",
-            "neutral": "😐 Netral",
-            "negative": "😡 Negatif"
-        }
-        
-        # Menyusun dictionary confidences untuk UI Progress Bar Gradio
-        confidences = {label_mapping[res['label']]: res['score'] for res in results}
-        
-        # Mencari sentimen dengan skor tertinggi
-        top_label = max(confidences, key=confidences.get)
-        
-        return f"### Kesimpulan: Teks bersentimen **{top_label}**", confidences
-        
-    except Exception as e:
-        return f"⚠️ Terjadi kesalahan: {str(e)}", {}
-
-# 2. ANTARMUKA GRADIO
-with gr.Blocks(theme=gr.themes.Soft()) as demo:
-    gr.Markdown("""
-    <h1 style='text-align: center;'>💬 AI Sentiment Monitoring System</h1>
-    <p style='text-align: center;'>Sistem NLP cerdas untuk mendeteksi emosi dan opini publik dari teks atau ulasan secara <i>real-time</i>.</p>
-    """)
-    
-    with gr.Row():
-        with gr.Column(scale=2):
-            # Input teks
-            inp_text = gr.Textbox(
-                label="📝 Masukkan Teks / Opini (Disarankan Bahasa Inggris)", 
-                placeholder="Contoh: The new product feature is absolutely amazing, but the customer service was terrible!",
-                lines=4
-            )
-            btn = gr.Button("🔍 Analisis Sentimen", variant="primary")
-            
-        with gr.Column(scale=1):
-            # Output hasil
-            out_kesimpulan = gr.Markdown(label="Kesimpulan AI")
-            out_label = gr.Label(label="📊 Distribusi Emosi (Probabilitas)")
-            
-    # Menambahkan beberapa contoh agar pengunjung GitHub/HF Anda bisa langsung mencoba
-    gr.Examples(
-        examples=[
-            "I absolutely love the new design, it's so intuitive and fast!",
-            "The delivery was late and the packaging was completely ruined. Very disappointed.",
-            "I bought this phone yesterday. The screen size is 6 inches and it comes with a charger.",
-            "It's okay I guess. Not the best, but definitely not the worst."
-        ],
-        inputs=inp_text
-    )
-            
-    # Hubungkan tombol
-    btn.click(fn=analyze_sentiment, inputs=inp_text, outputs=[out_kesimpulan, out_label])
-
-if __name__ == "__main__":
+import gradio as gr
+from transformers import pipeline
+
+# 1. LOAD MODEL NLP MULTILINGUAL
+print("🚀 Sedang memuat model NLP (Multilingual)...")
+# Menggunakan model DistilBERT Multilingual yang mendukung Bahasa Indonesia, Inggris, dan bahasa lainnya
+sentiment_pipeline = pipeline(
+    "text-classification", 
+    model="lxyuan/distilbert-base-multilingual-cased-sentiments-student", 
+    top_k=None
+)
+
+def analyze_sentiment(text):
+    if not text.strip():
+        return "⚠️ Teks kosong", {"😐 Netral": 1.0}
+    
+    try:
+        # AI memproses kalimat
+        results = sentiment_pipeline(text)[0]
+        
+        # Mapping label dari model (positive/neutral/negative) ke visualisasi UI bahasa Indonesia
+        label_mapping = {
+            "positive": "😊 Positif",
+            "neutral": "😐 Netral",
+            "negative": "😡 Negatif"
+        }
+        
+        # Menyusun dictionary confidences untuk UI Progress Bar Gradio
+        # Jika model mengeluarkan label lain, kita gunakan label aslinya
+        confidences = {label_mapping.get(res['label'], res['label'].capitalize()): res['score'] for res in results}
+        
+        # Mencari sentimen dengan skor tertinggi
+        top_label = max(confidences, key=confidences.get)
+        
+        return f"### Kesimpulan: Teks bersentimen **{top_label}**", confidences
+        
+    except Exception as e:
+        return f"⚠️ Terjadi kesalahan: {str(e)}", {}
+
+# 2. ANTARMUKA GRADIO
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.Markdown("""
+    <h1 style='text-align: center;'>💬 AI Sentiment Monitoring System</h1>
+    <p style='text-align: center;'>Sistem NLP cerdas untuk mendeteksi emosi dan opini publik dari teks atau ulasan dalam <b>Bahasa Indonesia</b> maupun <b>Bahasa Inggris</b>.</p>
+    """)
+    
+    with gr.Row():
+        with gr.Column(scale=2):
+            # Input teks sudah diperbarui untuk mengundang dua bahasa
+            inp_text = gr.Textbox(
+                label="📝 Masukkan Teks / Opini (ID / EN)", 
+                placeholder="Contoh: Fitur barunya sangat keren dan cepat, tapi sayang CS-nya sangat lambat membalas!",
+                lines=4
+            )
+            btn = gr.Button("🔍 Analisis Sentimen", variant="primary")
+            
+        with gr.Column(scale=1):
+            # Output hasil
+            out_kesimpulan = gr.Markdown(label="Kesimpulan AI")
+            out_label = gr.Label(label="📊 Distribusi Emosi (Probabilitas)")
+            
+    # Menambahkan contoh bilingual (campuran bahasa) agar pengunjung bisa langsung menguji kehebatannya
+    gr.Examples(
+        examples=[
+            "Pelayanannya sangat buruk, pesanan saya telat 3 hari dan kurirnya tidak ramah. Kecewa berat!",
+            "I absolutely love the new design, it's so intuitive and fast!",
+            "Harga produk ini lumayan mahal, tapi kualitasnya standar saja. Not bad lah.",
+            "The delivery was late but the product is okay."
+        ],
+        inputs=inp_text
+    )
+            
+    # Hubungkan tombol
+    btn.click(fn=analyze_sentiment, inputs=inp_text, outputs=[out_kesimpulan, out_label])
+
+if __name__ == "__main__":
     demo.launch()