Update app.py

app.py

@@ -43,12 +43,19 @@ def convert_audio_to_wav(audio_file):
         return None
 
 def evaluate_asr_accuracy(transcription, reference):
-    """Calculate WER and CER metrics"""
-    transformation = jiwer.Compose([
-        jiwer.ToLowerCase(), jiwer.RemovePunctuation(), jiwer.RemoveMultipleSpaces()
-    ])
-    wer = jiwer.wer(reference, transcription, truth_transform=transformation, hypothesis_transform=transformation)
-    return 1 - wer, 1 - jiwer.cer(reference, transcription, truth_transform=transformation)
+    """Simplified evaluation with direct string processing"""
+    # 简单预处理
+    ref_processed = reference.lower().strip()
+    hyp_processed = transcription.lower().strip()
+    
+    if not ref_processed:  # 参考文本为空
+        return 0.0, 0.0
+    
+    # 计算WER和CER
+    wer = jiwer.wer(ref_processed, hyp_processed)
+    cer = jiwer.cer(ref_processed, hyp_processed)
+    
+    return 1 - wer, 1 - cer
 
 # Cached dataset loading with audio decoding
 @st.cache_data(show_spinner=False)