Update app.py

app.py

@@ -7,13 +7,17 @@ from gtts import gTTS
 import io
 from pydub import AudioSegment
 import time
-import re
+import pronouncing
+import epitran
 
 # Create audio directory if it doesn't exist
 if not os.path.exists('audio'):
     os.makedirs('audio')
 
-# Function to upload file to server
+# Initialize the epitran object for English
+epi = epitran.Epitran('eng-Latn')
+
+# Step 2: Create pronunciation audio for incorrect words
 def upfilepath(local_filename):
     ts = time.time()
     upload_url = f"https://mr2along-speech-recognize.hf.space/gradio_api/upload?upload_id={ts}"
@@ -78,22 +82,35 @@ def create_pronunciation_audio(word):
     except Exception as e:
         return f"Failed to create pronunciation audio: {e}"
 
-# Function to respell words phonetically
-def phonetic_respelling(word):
-    word = word.lower()
-    respelling = word
-    # Simplified phonetic transformations
-    respelling = re.sub(r'th', 'th', respelling)  # 'th' as in 'the'
-    respelling = re.sub(r'[aeiou]', lambda m: f'{m.group(0)}', respelling)  # Simplify vowels
-    respelling = re.sub(r'c', 'k', respelling)  # 'c' sounds like 'k'
-    respelling = re.sub(r'ph', 'f', respelling)  # 'ph' as 'f'
-    respelling = re.sub(r'ch', 'ch', respelling)  # 'ch' as in 'church'
-    respelling = re.sub(r'qu', 'kw', respelling)  # 'qu' as 'kw'
-    respelling = re.sub(r'sh', 'sh', respelling)  # 'sh' as in 'shoe'
-    respelling = re.sub(r'[^a-z]', '', respelling)  # Remove non-alphabet characters
+# Function for phonetic respelling
+def phonetic_respelling(sentence):
+    words = sentence.split()
+    respelled = []
+    
+    for word in words:
+        # Find close matches for each word
+        close_matches = pronouncing.search(word)
+        if close_matches:
+            # Get the first close match
+            closest_word = close_matches[0]
+            respelled.append(pronouncing.phones_for_word(closest_word)[0])  # Use phonemes for the closest match
+        else:
+            respelled.append(word)
+
+    # Convert phonemes to respelling
+    respelling = ' '.join(respelled)
+    
+    # Replace phonemes with common respellings
+    respelling = respelling.replace('ˈ', '').replace('ˌ', '').replace('ː', '')  # Clean up phoneme symbols
+    respelling = respelling.replace('ɑ', 'a').replace('ə', 'uh').replace('ɪ', 'i').replace('ʊ', 'u')  # Sample conversions
+    
     return respelling
 
-# Step 3: Compare the transcribed text with the input paragraph and add phonetic respelling
+# Function for IPA transcription
+def ipa_transcription(sentence):
+    return epi.transliterate(sentence)
+
+# Step 3: Compare the transcribed text with the input paragraph
 def compare_texts(reference_text, transcribed_text):
     reference_words = reference_text.split()
     transcribed_words = transcribed_text.split()
@@ -115,11 +132,9 @@ def compare_texts(reference_text, transcribed_text):
 
     html_output += f"<strong>Quality Score:</strong> {similarity_score}%<br>"
     html_output += f"<strong>Transcribed Text:</strong> {transcribed_text}<br>"
-    
-    # Add phonetic respelling for the entire sentence
-    phonetic_transcription = " ".join([phonetic_respelling(word) for word in transcribed_words])
-    html_output += f"<strong>Phonetic Respelling:</strong> {phonetic_transcription}<br>"
-
+    html_output += f"<strong>Input Sentence:</strong> {reference_text}<br>"
+    html_output += f"<strong>Phonetic Respelling:</strong> {phonetic_respelling(reference_text)}<br>"
+    html_output += f"<strong>IPA Transcription:</strong> {ipa_transcription(reference_text)}<br>"
     html_output += "<strong>Word Score List:</strong><br>"
 
     # Generate colored word score list
@@ -145,12 +160,11 @@ def compare_texts(reference_text, transcribed_text):
         for word, audio in incorrect_words_audios:
             suggestion = difflib.get_close_matches(word, reference_words, n=1)
             suggestion_text = f" (Did you mean: <em>{suggestion[0]}</em>?)" if suggestion else ""
-            up_audio=upfilepath(audio)
-            audio_src=f"https://mr2along-speech-recognize.hf.space/gradio_api/file={up_audio}"
+            up_audio = upfilepath(audio)
+            audio_src = f"https://mr2along-speech-recognize.hf.space/gradio_api/file={up_audio}"
             html_output += f'{word}: '
             html_output += f'<audio controls><source src="{audio_src}" type="audio/mpeg">Your browser does not support the audio tag.</audio>{suggestion_text}<br>'
 
-    # Return the final result with phonetic respelling
     return [html_output]
 
 # Step 4: Text-to-Speech Function