Update app.py

app.py

@@ -18,14 +18,18 @@ except:
     pass
 
 import gradio as gr
+import spacy
 from fastcoref import FCoref
 from deep_translator import GoogleTranslator
 
-# 3. 初始化輕量級消解模型
+# 3. 初始化 Spacy NLP 詞性解析器與指代模型
+print("🚀 [System] 正在初始化 NLP 詞性解析器...")
+nlp = spacy.load("en_core_web_sm")
+
 print("🚀 [System] 正在初始化 F-Coref 核心大腦...")
 try:
     model = FCoref(model_name_or_path='biu-nlp/f-coref', device='cpu')
-    print("✅ [System] 模型加載成功！")
+    print("✅ [System] 所有模型加載成功！")
 except:
     model = FCoref('biu-nlp/f-coref', device='cpu')
     print("✅ [System] 備用路徑加載成功！")
@@ -52,30 +56,28 @@ def coref_learning_pipeline(user_input):
         # C. 欄位一：生成「英翻中結果」
         translation_text = GoogleTranslator(source='en', target='zh-TW').translate(working_text)
         
-        # D. 欄位二：生成妳要的【AI 智慧單字本】
+        # D. 💥 欄位二：全新重構【AI 智慧單字本】（利用詞性精準抓取名詞，消滅人名）
         vocab_output = ""
-        if clusters:
-            extracted_words = set()
-            for cluster in clusters:
-                for item in cluster:
-                    # 清理冠詞與代名詞，只留下核心實體名詞
-                    clean_word = item.lower().replace("the ", "").replace("a ", "").replace("her ", "").replace("his ", "").strip()
-                    if len(clean_word) > 2 and clean_word not in ['she', 'he', 'him', 'her', 'it', 'they', 'i', 'you']:
-                        extracted_words.add(clean_word)
-            
-            if extracted_words:
-                for word in sorted(extracted_words):
-                    try:
-                        # 翻成繁體中文
-                        word_zh = GoogleTranslator(source='en', target='zh-TW').translate(word)
-                        # 格式化成妳要的：【Farmer (農夫)】樣式
-                        vocab_output += f"🔸 {word.capitalize()} ➔ {word_zh}\n"
-                    except:
-                        pass
-            else:
-                vocab_output = "ℹ️ 本句的核心實體多為代名詞，未偵測到適合學習的高階名詞。"
+        doc = nlp(working_text)
+        extracted_words = set()
+        
+        for token in doc:
+            # 只抓普通名詞 (NOUN)，排除專有名詞/人名 (PROPN) 與代名詞 (PRON)
+            if token.pos_ == "NOUN" and len(token.text) > 2:
+                # 統一轉成單數原型原型，畫面更漂亮（例如把 lambs 變成 lamb）
+                extracted_words.add(token.lemma_.lower())
+        
+        if extracted_words:
+            for word in sorted(extracted_words):
+                try:
+                    # 翻成繁體中文
+                    word_zh = GoogleTranslator(source='en', target='zh-TW').translate(word)
+                    # 輸出格式：【Farmer ➔ 農夫】
+                    vocab_output += f"🔸 {word.capitalize()} ➔ {word_zh}\n"
+                except:
+                    pass
         else:
-            vocab_output = "🔍 未偵測到明確的指代關係，嘗試輸入更長的段落以提取單字！"
+            vocab_output = "ℹ️ 未偵測到適合學習的核心英文單字。"
 
         # E. 欄位三：生成「AI 語意共指報告」
         report_text = f"✨ 系統狀態：{mode_notice}\n"
@@ -100,26 +102,25 @@ def coref_learning_pipeline(user_input):
     except Exception as e:
         return f"錯誤: {str(e)}", "無法整合單字", f"運行異常: {str(e)}"
 
-# 5. 精美 UI 介面設計（利用 Blocks 切分獨立區塊）
+# 5. 精美 UI 介面設計
 with gr.Blocks(theme=gr.themes.Soft(primary_hue="blue", secondary_hue="indigo")) as demo:
     gr.Markdown("# 🤖 AI 跨語言智慧語意學習終端")
-    gr.Markdown("### 🚀 專題亮點：結合核心指代消解 (Coreference Resolution) 與動態單字卡生成技術")
+    gr.Markdown("### 🚀 專題亮點：結合核心指代消解 (Coreference Resolution) 與 NLP 智慧名詞提取技術")
     
     with gr.Row():
         with gr.Column(scale=1):
             txt_input = gr.Textbox(
                 label="📥 請輸入中文或英文段落 (Input Text)", 
-                placeholder="例如：Mary is a Farmer. Mary had a little lamb.", 
+                placeholder="例如：Mary is a farmer. Mary has a little lamb.", 
                 lines=5
             )
             btn_submit = gr.Button("🔥 執行多維度 AI 語意解析", variant="primary")
             
         with gr.Column(scale=1):
-            # 這三個是各自獨立的漂亮框框！
             out_translation = gr.Textbox(label="📖 完整文本翻譯 (Translation)", lines=3)
-            # 這裡就是妳要的專屬單字本！
-            out_vocab = gr.Textbox(label="📚 AI 智慧單字本 (Vocabulary Booklet)", lines=4)
-            out_report = gr.Textbox(label="🎯 AI 語意消解報告 (Coreference Report)", lines=6)
+            # 這裡就是真正的智慧單字本！
+            out_vocab = gr.Textbox(label="📚 AI 智慧單字本 (Vocabulary Booklet)", lines=5)
+            out_report = gr.Textbox(label="🎯 AI 語意消解報告 (Coreference Report)", lines=5)
             
     btn_submit.click(
         fn=coref_learning_pipeline, 
@@ -127,13 +128,5 @@ with gr.Blocks(theme=gr.themes.Soft(primary_hue="blue", secondary_hue="indigo"))
         outputs=[out_translation, out_vocab, out_report]
     )
 
-    gr.Examples(
-        examples=[
-            ["Mary is a Farmer. Mary has a little lamb. John is a Florist. John has a flower shop."],
-            ["張醫生正在動手術。她很冷靜。她的助手在旁邊遞器材，他看起來很緊張。"]
-        ],
-        inputs=txt_input
-    )
-
 if __name__ == "__main__":
     demo.launch()