algo update

app/matcher.py

@@ -18,29 +18,9 @@ def sequence_match(a, b):
     return difflib.SequenceMatcher(None, a, b).ratio()
 
 
-# Main function to compare texts with percentage match
-def compare_texts(text1, text2):
-    """
-    Compares two texts using phonetic matching and sequence matching,
-    returning a percentage match score.
-    """
-    words1 = text1.lower().split()
-    words2 = text2.lower().split()
-
-    total_matches = len(words1)
-    mismatches = 0
-
-    for word1, word2 in zip(words1, words2):
-        if word1 != word2:
-            mismatches += 1
-            if phonetic_match(word1, word2) < 80:
-                # Use sequence matching only if phonetic is low
-                if sequence_match(word1, word2) < 0.8:
-                    mismatches += 1  # Penalty for bad sequence match
-
-    accuracy = 1 - (mismatches / total_matches)
-    return accuracy * 100  # Convert to percentage
 
 
 def match(original, transcription):
-    return compare_texts(original, transcription)
+    sequence = sequence_match(original, transcription)
+    phonetic = phonetic_match(original, transcription)
+    return sequence, phonetic

app/routers/V1/voice/voice_router.py

@@ -6,6 +6,7 @@ import os
 from app.transcriber import get_transcription
 from app.matcher import match
 from app.mfcc import mfcc_similarty_check
+from app.string_processor import process_text
 
 
 """ initialize the router """
@@ -40,7 +41,6 @@ async def transcribe_audio(
         with open(filename_original, "wb") as buffer:
             buffer.write(original_bytes)
 
-
         # Read file bytes
         recorded_bytes = await recorded.read()
         filename_recorded = f"audio_{int(time.time())}_recorded.wav"
@@ -49,19 +49,21 @@ async def transcribe_audio(
         with open(filename_recorded, "wb") as buffer:
             buffer.write(recorded_bytes)
 
-
         try:
             text = get_transcription(filename_recorded)
-            percent = match(matcher_text, text.strip())
-            Euclidean, Cosine = mfcc_similarty_check(filename_original, filename_recorded)
+            sequence, phonetic = match(matcher_text, process_text(text))
+            Euclidean, Cosine = mfcc_similarty_check(
+                filename_original, filename_recorded
+            )
             return JSONResponse(
-                    {
-                        "transcription": text,
-                        "percent": percent,
-                        "Cosine": Cosine,
-                        "Euclidean": Euclidean,
-                    }
-                )
+                {
+                    "transcription": text,
+                    "sequence": sequence,
+                    "phonetic": phonetic,
+                    "Cosine": Cosine,
+                    "Euclidean": Euclidean,
+                }
+            )
         finally:
             # Clean up the temporary file
             os.remove(filename_original)

app/string_processor.py

@@ -0,0 +1,18 @@
+import string
+import re
+
+
+def process_text(text):
+    # Step 1: Strip whitespace from both ends
+    text = text.strip()
+
+    # Step 2: Remove all punctuation (including full stops and commas)
+    text = text.translate(str.maketrans("", "", string.punctuation))
+
+    # Step 3: Extract sentences (assuming you want to keep the text as a whole sentence)
+    sentences = re.split(r"(?<=[.!?]) +", text)
+
+    # Combine the sentences back into a single string without punctuation
+    processed_text = " ".join(sentences)
+
+    return processed_text