save outputs

README.md

@@ -6,4 +6,9 @@
 
 ### To run the code
 `pip install -r requirements.txt`
-`python whisper.py`
+
+執行所有在data目錄下的語音檔
+`python whisper.py`
+
+執行單個語音檔
+`python whisper.py --file 語音檔名`

whisper.py

@@ -5,6 +5,7 @@ from dotenv import load_dotenv
 import tiktoken
 from pypinyin import pinyin, Style
 import jieba
+from datetime import datetime
 
 load_dotenv('environment.env')
 client = OpenAI()
@@ -31,6 +32,23 @@ def transcribe(audio_file):
     except Exception as e:
         print(f"轉錄時發生錯誤：{str(e)}")
         return None
+    
+def save_output(file_name, raw_transcript, corrected_transcript):
+    output_dir = "output"
+    if not os.path.exists(output_dir):
+        os.makedirs(output_dir)
+    
+    output_file = os.path.join(output_dir, "transcription_results.txt")
+    
+    with open(output_file, "a", encoding="utf-8") as f:
+        f.write(f"\n{'='*50}\n")
+        f.write(f"文件名: {file_name}\n")
+        f.write(f"處理時間: {datetime.now().strftime('%Y-%m-%d %H:%M:%S')}\n\n")
+        f.write("原始轉錄:\n")
+        f.write(f"{raw_transcript}\n\n")
+        f.write("修正後的轉錄:\n")
+        f.write(f"{corrected_transcript}\n")
+
 
 def process_audio_file(file_path):
     try:
@@ -51,9 +69,13 @@ def process_audio_file(file_path):
             print("\n修正後的轉錄：")
             print(corrected_transcript)
 
+            # 保存輸出結果
+            save_output(os.path.basename(file_path), raw_transcript, corrected_transcript)
+
     except Exception as e:
         print(f"處理文件 {os.path.basename(file_path)} 時發生錯誤：{str(e)}")
 
+
 def process_folder(folder_path):
     processed_files = 0
 
@@ -106,6 +128,7 @@ error_correction = {
     "看": "碳"
 }
 
+
 def fuzzy_correct_chinese(text, correct_terms):
     words = jieba.cut(text)
     corrected_words = []
@@ -121,6 +144,7 @@ def fuzzy_correct_chinese(text, correct_terms):
                 corrected_words.append(word)
     return ' '.join(corrected_words)
 
+
 def post_process_transcript(transcript, temperature=0):
     correct_terms = ["碳", "溫室氣體", "碳排放", "排放", "碳管理", "管理", "碳盤查", "盤查", "碳權交易", "碳費",
                      "碳權", "碳足跡", "足跡", "淨零排放", "零排放", "排放", "淨零",
@@ -143,6 +167,7 @@ def post_process_transcript(transcript, temperature=0):
 
     return response.choices[0].message.content
 
+
 def main():
     parser = argparse.ArgumentParser(description="處理音頻文件使用 Whisper")
     parser.add_argument("--file", help="要處理的單個音頻文件的路徑")