Updated with text processing for better comparison

app.py

@@ -2,6 +2,8 @@ from difflib import Differ
 
 import gradio as gr
 
+import string
+
 import torch
 from transformers import (
     AutoModelForSpeechSeq2Seq,
@@ -55,6 +57,13 @@ pipe = pipeline(
 )
 
 
+def prepare_text_for_comparison(text_to_clean: str):
+
+    text_to_clean = text_to_clean.translate(str.maketrans("", "", string.punctuation))
+
+    return text_to_clean.casefold()
+
+
 def diff_texts(diction_text: str, audio_input: str):
     d = Differ()
 
@@ -76,7 +85,10 @@ def calc_score(diff_texts: list) -> float:
 
 def transcribe_audio(diction_text, audio):
     result = pipe(audio)
-    diff_text = diff_texts(diction_text, result["text"])
+
+    cleaned_result = prepare_text_for_comparison(result["text"])
+    cleaned_diction_text = prepare_text_for_comparison(diction_text)
+    diff_text = diff_texts(cleaned_diction_text, cleaned_result)
 
     score = calc_score(diff_text)