segment database by language (#9)

- segment database by language (0b24e727e18a34154af419cf6c0e8fd5746db855)


Co-authored-by: Hunter S <[email protected]>

app.py

@@ -11,13 +11,15 @@ import json
 import tempfile
 import uuid
 
-''' LOCAL INITIALIZATION - ONLY USE ON YOUR OWN DEVICE 
+# LOCAL INITIALIZATION - ONLY USE ON YOUR OWN DEVICE 
+'''
 os.chdir(os.path.dirname(os.path.abspath(__file__)))
 cred = credentials.Certificate("serviceAccountKey.json")
 '''
 # Deployed Initialization
 firebase_config = json.loads(os.environ.get('firebase_creds'))
 cred = credentials.Certificate(firebase_config)
+
 firebase_admin.initialize_app(cred, {
     "storageBucket": "amis-asr-corrections-dem-8cf3d.firebasestorage.app"
 })
@@ -26,6 +28,7 @@ bucket = storage.bucket()
 
 # Load the ASR model and processor
 MODEL_NAME = "eleferrand/xlsr53_Amis"
+lang = "ami"
 processor = Wav2Vec2Processor.from_pretrained(MODEL_NAME)
 model = AutoModelForCTC.from_pretrained(MODEL_NAME)
 
@@ -54,14 +57,14 @@ def store_correction(original_transcription, corrected_transcription, audio_file
         
         # If an audio file is provided, upload it to Firebase Storage
         if audio_file and os.path.exists(audio_file):
-            audio, sr = librosa.load(audio_file, sr=16000)
+            audio, sr = librosa.load(audio_file, sr=44100)
             duration = librosa.get_duration(y=audio, sr=sr)
             file_size = os.path.getsize(audio_file)
             audio_metadata = {'duration': duration, 'file_size': file_size}
             
             # Generate a unique identifier for the audio file
             unique_id = str(uuid.uuid4())
-            destination_path = f"audio/{unique_id}.mp3"
+            destination_path = f"audio/{lang}/{unique_id}.wav"
             
             # Create a blob and upload the file
             blob = bucket.blob(destination_path)
@@ -71,18 +74,24 @@ def store_correction(original_transcription, corrected_transcription, audio_file
             audio_file_url = blob.generate_signed_url(expiration=timedelta(hours=1))
         
         combined_data = {
-            'original_text': original_transcription,
-            'corrected_text': corrected_transcription,
-            'timestamp': datetime.now().isoformat(),
-            'audio_metadata': audio_metadata,
-            'audio_file_url': audio_file_url,
-            'model_name': MODEL_NAME,
+            'transcription_info': {
+                'original_text': original_transcription,
+                'corrected_text': corrected_transcription,
+                'language': lang,
+            },
+            'audio_data': {
+                'audio_metadata': audio_metadata,
+                'audio_file_url': audio_file_url,
+            },
             'user_info': {
                 'native_amis_speaker': native_speaker,
                 'age': age
-            }
+            },
+            'timestamp': datetime.now().isoformat(),
+            'model_name': MODEL_NAME
         }
-        db.collection('transcriptions').add(combined_data)
+        # Save data to a collection for that language 
+        db.collection('amis_transcriptions').add(combined_data)
         return "校正保存成功! (Correction saved successfully!)"
     except Exception as e:
         return f"保存失败: {e} (Error saving correction: {e})"