Update app.py

app.py

@@ -52,16 +52,24 @@ def extract_text_from_image(filepath: str, languages: List[str]):
     lang_str = '+'.join(languages)  # Join languages for Tesseract
     return pytesseract.image_to_string(image=image, lang=lang_str)
 
-# Function to get embeddings for text using SentenceTransformer
-def get_embedding(text):
-    return model1.encode(text, convert_to_tensor=True)
 
-# Calculate similarity between two texts focusing on sequence
-def calculate_similarity(text1, text2):
-    embedding1 = get_embedding(text1)
-    embedding2 = get_embedding(text2)
-    similarity = util.pytorch_cos_sim(embedding1, embedding2)
-    return similarity.item()
+
+import Levenshtein
+
+# Calculate Levenshtein Distance
+def calculate_levenshtein_distance(text1, text2):
+    return Levenshtein.distance(text1, text2)
+
+# Use Levenshtein distance to refine similarity
+def calculate_similarity_with_levenshtein(text1, text2):
+    similarity = calculate_similarity(text1, text2)  # From embeddings
+    edit_distance = calculate_levenshtein_distance(text1, text2)
+    normalized_distance = 1 - (edit_distance / max(len(text1), len(text2)))
+    
+    # Combine similarity score with normalized Levenshtein distance
+    final_similarity = (similarity + normalized_distance) / 2
+    return final_similarity
+
 
 # Assign badges based on the grade
 def assign_badge(grade):
@@ -74,21 +82,6 @@ def assign_badge(grade):
     else:
         return "Keep Improving Badge 💪"
 
-# Generate sequence-based feedback
-def generate_sequence_feedback(student_answer, model_answer):
-    student_words = student_answer.split()
-    model_words = model_answer.split()
-    common_sequences = []
-    
-    for i in range(len(student_words)):
-        sequence = ' '.join(student_words[i:i+3])  # Compare sequences of 3 words at a time
-        if sequence in model_answer:
-            common_sequences.append(f"**{sequence}**")
-        else:
-            common_sequences.append(sequence)
-    
-    highlighted_feedback = ' '.join(common_sequences)
-    return highlighted_feedback
 
 # Categorize feedback into clarity, completeness, and accuracy
 def detailed_feedback(similarity_score):
@@ -117,7 +110,7 @@ def get_grade(similarity_score):
 # Function to evaluate student's answer by comparing it to a model answer
 def evaluate_answer(image, languages, model_answer):
     student_answer = extract_text_from_image(image, languages)
-    similarity_score = calculate_similarity(student_answer, model_answer)
+    similarity_score = calculate_similarity_with_levenshtein(student_answer, model_answer)
     grade = get_grade(similarity_score)
     feedback = f"Student's answer: {student_answer}\nTeacher's answer: {model_answer}"
     visual_feedback = generate_sequence_feedback(student_answer, model_answer)