Merge pull request #260 from aidancrowther/add-detection-confidence

Add confidence percentage to language detection

CHANGELOG.md

@@ -6,6 +6,7 @@ Unreleased
 
 ### Added
 
+  - Added detection confidence to langauge detection endpoint
   - Set mel generation to adjust n_dims automatically to match the loaded model
 
 [1.6.0] (2024-10-06)

app/faster_whisper/core.py

@@ -70,8 +70,9 @@ def language_detection(audio):
     with model_lock:
         segments, info = model.transcribe(audio, beam_size=5)
         detected_lang_code = info.language
+        detected_language_confidence = info.language_probability
 
-    return detected_lang_code
+    return detected_lang_code, detected_language_confidence
 
 
 def write_result(result: dict, file: BinaryIO, output: Union[str, None]):

app/openai_whisper/core.py

@@ -55,7 +55,7 @@ def language_detection(audio):
         _, probs = model.detect_language(mel)
     detected_lang_code = max(probs, key=probs.get)
 
-    return detected_lang_code
+    return detected_lang_code, probs[max(probs)]
 
 
 def write_result(result: dict, file: BinaryIO, output: Union[str, None]):

app/webservice.py

@@ -89,8 +89,8 @@ async def detect_language(
     audio_file: UploadFile = File(...),  # noqa: B008
     encode: bool = Query(default=True, description="Encode audio first through FFmpeg"),
 ):
-    detected_lang_code = language_detection(load_audio(audio_file.file, encode))
-    return {"detected_language": tokenizer.LANGUAGES[detected_lang_code], "language_code": detected_lang_code}
+    detected_lang_code, confidence = language_detection(load_audio(audio_file.file, encode))
+    return {"detected_language": tokenizer.LANGUAGES[detected_lang_code], "language_code": detected_lang_code, "confidence": confidence}
 
 
 def load_audio(file: BinaryIO, encode=True, sr: int = SAMPLE_RATE):