Update app.py

app.py

@@ -1,7 +1,10 @@
 import streamlit as st
 from transformers import pipeline
-import textwrap
 import re
+import nltk
+
+nltk.download('punkt')
+from nltk.tokenize import sent_tokenize
 
 st.title('Hugging Face BERT Summarizer')
 
@@ -20,14 +23,16 @@ keywords = st.text_input("Enter keywords (comma-separated)")
 scale_percentage = st.sidebar.slider('Scale %', min_value=1, max_value=100, value=50)
 
 # Add slider for the chunk size
-chunk_size = st.sidebar.slider('Chunk size', min_value=100, max_value=1000, value=500)
+chunk_size = st.sidebar.slider('Chunk size (words)', min_value=100, max_value=1000, value=500)
 
 if uploaded_file is not None and keywords:
     user_input = uploaded_file.read().decode('utf-8')
     keywords = [keyword.strip() for keyword in keywords.split(",")]
 
+    # Split text into sentences
+    sentences = sent_tokenize(user_input)
+    
     # Filter sentences based on keywords
-    sentences = re.split(r'(?<=[^A-Z].[.?]) +(?=[A-Z])', user_input)
     filtered_sentences = [sentence for sentence in sentences if any(keyword.lower() in sentence.lower() for keyword in keywords)]
     filtered_text = ' '.join(filtered_sentences)
 
@@ -35,16 +40,17 @@ if uploaded_file is not None and keywords:
         summarizer = pipeline('summarization', model=model)
         summarized_text = ""
 
-        # Split the filtered text into chunks of approximately the selected chunk size each
-        chunks = textwrap.wrap(filtered_text, chunk_size)
+        # Split filtered text into chunks by words
+        words = filtered_text.split()
+        chunks = [' '.join(words[i:i+chunk_size]) for i in range(0, len(words), chunk_size)]
         
         # Summarize each chunk
         for chunk in chunks:
             chunk_length = len(chunk.split())
-            min_length_percentage = max(scale_percentage - 10, 1)  # Ensure min_length_percentage is not less than 1
-            max_length_percentage = min(scale_percentage + 10, 100)  # Ensure max_length_percentage is not more than 100
-            min_length = max(int(chunk_length * min_length_percentage / 100), 1)  # Calculate min_length based on the percentage of the chunk length
-            max_length = int(chunk_length * max_length_percentage / 100)  # Calculate max_length based on the percentage of the chunk length
+            min_length_percentage = max(scale_percentage - 10, 1)
+            max_length_percentage = min(scale_percentage + 10, 100)
+            min_length = max(int(chunk_length * min_length_percentage / 100), 1)
+            max_length = int(chunk_length * max_length_percentage / 100)
             summarized = summarizer(chunk, max_length=max_length, min_length=min_length, do_sample=False)
             summarized_text += summarized[0]['summary_text'] + " "