add function to check for hallucination stats

app.py

@@ -1,6 +1,7 @@
 import streamlit as st
 import pandas as pd
 import ast
+import re
 import PyPDF2
 from PyPDF2 import PdfReader
 import openreview
@@ -20,6 +21,16 @@ client_openai = OpenAI(api_key=openai_api_key)
 anthropic_api_key = os.environ["ANTHROPIC_API_KEY"]
 client_anthropic = Anthropic(api_key=anthropic_api_key)
 
+def parse_quotes(input_string, pdf_text):
+    # Find all matches of <quote>...</quote> and extract the content between the tags
+    matches = re.findall(r'<quote>(.*?)</quote>', input_string)
+    
+    count = len(matches)
+    extracted_texts = matches
+    match_count = sum(1 for text in extracted_texts if text in pdf_text)
+    
+    return count, match_count
+
 # load manual (human) annotations
 def load_annotations():
     path = './annotations_8_26.csv'
@@ -451,6 +462,9 @@ if user_input:
         best_feedback = create_feedback(review, pdf_text, agent_prompt, model)
     
     revised_feedback = critic(review, best_feedback, pdf_text, critic_prompt, model)
+    
+    count, match_count = parse_quotes(revised_feedback, pdf_text)
+    
     revised_feedback = revised_feedback.replace("<quote>", "'").replace("</quote>", "'")
 
     st.title(f'Review feedback')
@@ -489,6 +503,7 @@ if user_input:
         # Column 4: Display revised feedback (from critic)
         with col4:
             st.subheader('Revised Feedback')
+            st.write(f"Quotes found verbatim in pdf text: {match_count}/{count}")
             st.write(revised_feedback)
     else:
 
@@ -512,6 +527,7 @@ if user_input:
         # Column 3: Display revised feedback (from critic)
         with col3:
             st.subheader('Revised Feedback')
+            st.write(f"Quotes found verbatim in pdf text: {match_count}/{count}")
             st.write(revised_feedback)
 
 else: