Update app.py

app.py

@@ -6,6 +6,7 @@ from prophet import Prophet
 import matplotlib.pyplot as plt
 import gradio as gr
 
+# model
 embedder = SentenceTransformer("sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2")
 sentiment_model = pipeline(
     "text-classification",
@@ -13,6 +14,7 @@ sentiment_model = pipeline(
     tokenizer="uer/roberta-base-finetuned-dianping-chinese"
 )
 
+#main
 def full_pipeline(file, num_clusters):
     df = pd.read_csv(file)
 
@@ -21,12 +23,13 @@ def full_pipeline(file, num_clusters):
     if "timestamp" not in df.columns:
         return "❌ 錯誤：CSV 檔案需包含 timestamp 欄位（例如新聞時間）"
 
+    #降維
     texts = df["text"].astype(str).tolist()
     embeddings = embedder.encode(texts, show_progress_bar=True)
     kmeans = KMeans(n_clusters=num_clusters, random_state=42)
     df["topic"] = kmeans.fit_predict(embeddings)
 
-
+    # 情緒分析
     sentiments = []
     for text in texts:
         try:
@@ -45,7 +48,7 @@ def full_pipeline(file, num_clusters):
         sentiments.append(sentiment)
     df["sentiment"] = sentiments
 
-
+    # 熱度預測
     df["timestamp"] = pd.to_datetime(df["timestamp"])
     topic0 = df[df["topic"] == 0]
     daily_counts = topic0.groupby(df["timestamp"].dt.date).size().reset_index(name="count")
@@ -60,7 +63,7 @@ def full_pipeline(file, num_clusters):
     forecast = m.predict(future)
     fig = m.plot(forecast)
 
-    # 儲存結果
+    #output
     output_csv = "/tmp/final_output.csv"
     output_img = "/tmp/forecast.png"
     df.to_csv(output_csv, index=False)
@@ -68,7 +71,7 @@ def full_pipeline(file, num_clusters):
 
     return output_csv, output_img
 
-# 3️⃣ Gradio 介面
+#gradio
 gr.Interface(
     fn=full_pipeline,
     inputs=[
@@ -79,6 +82,6 @@ gr.Interface(
         gr.File(label="結果 CSV（含 topic, sentiment）"),
         gr.Image(label="topic=0 熱度預測圖（Prophet）")
     ],
-    title=,"話題雷達"
+    title="話題雷達",
     description="自動分群、分析情緒，並預測熱度走勢（topic=0 為例）"
 ).launch()