add fillerscore

filler_count/filler_score.py

@@ -1,24 +1,76 @@
 import re
 import whisper
+from pydub import AudioSegment  # For accurate duration calculation
 
 def analyze_fillers(file_path: str, model_size: str = "base") -> dict:
+    """
+    Analyzes English filler words in audio with proper duration handling.
+    """
     try:
-        FILLER_WORDS = ["um", "uh", "hmm", "ah", "er", "eh", "like", "you know", "well"]
+        FILLER_WORDS = [
+            "um", "uh", "hmm", "ah", "er", "eh", 
+            "umm", "uhh", "mmm", "ahh", "err",
+            "like", "you know", "well", "so", "actually", "basically",
+            "right", "okay", "sort of", "kind of"
+        ]
         
+        # First get accurate duration using pydub
+        audio = AudioSegment.from_file(file_path)
+        duration = len(audio) / 1000  # Convert ms to seconds
+        
+        # Then run Whisper transcription
         model = whisper.load_model(model_size)
         result = model.transcribe(file_path, word_timestamps=False, fp16=False)
         transcript = result["text"]
         
-        pattern = r"\b(" + "|".join(FILLER_WORDS) + r")\b"
-        matches = re.findall(pattern, transcript.lower())
+        # Case-insensitive regex matching
+        pattern = r"(?<!\w)(" + "|".join(map(re.escape, FILLER_WORDS)) + r")(?!\w)"
+        matches = re.findall(pattern, transcript, re.IGNORECASE)
         
-        filler_counts = {filler: matches.count(filler) for filler in FILLER_WORDS}
+        # Count occurrences
+        filler_counts = {}
+        for word in matches:
+            key = word.lower()
+            filler_counts[key] = filler_counts.get(key, 0) + 1
         total_fillers = sum(filler_counts.values())
-    
+        
+        # Calculate rate per minute
+        filler_per_min = (total_fillers / duration) * 60 if duration > 0 else 0
+        
+        # Scoring
+        if total_fillers == 0:
+            filler_score = 100
+        elif filler_per_min < 1:
+            filler_score = 90
+        elif filler_per_min < 3:
+            filler_score = 80
+        elif filler_per_min < 5:
+            filler_score = 60
+        elif filler_per_min < 10:
+            filler_score = 40
+        else:
+            filler_score = 20
+        
+        # Generate insight
+        top_fillers = sorted(filler_counts.items(), key=lambda x: x[1], reverse=True)[:2]
+        
+        if total_fillers == 0:
+            insight = "Excellent! No filler words detected."
+        elif total_fillers <= 2:
+            insight = f"Minimal fillers ({total_fillers} total), mostly '{top_fillers[0][0]}'."
+        elif total_fillers <= 5:
+            examples = ", ".join(f"'{f[0]}'" for f in top_fillers)
+            insight = f"Moderate fillers ({total_fillers} total), mainly {examples}."
+        else:
+            examples = ", ".join(f"'{f[0]}'" for f in top_fillers)
+            insight = f"Excessive fillers ({total_fillers} total), dominated by {examples}."
+        
         return {
-            # "transcript": transcript,
-            "filler_counts": {k: v for k, v in filler_counts.items() if v > 0},
-            "total_fillers": total_fillers
+            "filler_counts": filler_counts,
+            "total_fillers": total_fillers,
+            "filler_score": filler_score,
+            "filler_rate_per_min": round(filler_per_min, 1),
         }
+        
     except Exception as e:
-        raise RuntimeError(f"Error during analysis: {str(e)}")
+        raise RuntimeError(f"Analysis failed: {str(e)}")