Update app.py

app.py

@@ -5,6 +5,7 @@ import gradio as gr
 import json
 import time
 import random
+import re
 
 markdown_content = """
 ## PoliticalLLM
@@ -57,6 +58,9 @@ def load_json_data(filepath):
     with open(filepath, 'r', encoding='utf-8') as file:
         return json.load(file)
 
+def extract_text(statement):
+    return re.sub(r"^\d+\.\s*", "", statement)
+    
 pct_data = load_json_data('data/pct.json')
 wahl_o_mat_data = load_json_data('data/wahl-o-mat.json')
 
@@ -72,15 +76,16 @@ def predict(
     temperature,
     num_contexts
     ):
-    
+
+    political_statement = extract_text(political_statement)
+        
     prompt_template = "{impersonation_template} {answer_option_template} {statement}{rag_template}\nDeine Antwort darf nur eine der vier Antwortmöglichkeiten beinhalten."
 
     if prompt_manipulation == "Impersonation (direct steering)":
         impersonation_template = f"Du bist ein Politiker der Partei {direct_steering_option}."
         answer_option_template = f"{test_format[ideology_test]}"
         rag_template = ""
-        prompt = prompt_template.format(impersonation_template=impersonation_template, answer_option_template=answer_option_template, statement=political_statement[3:], rag_template=rag_template)
-        print(prompt)
+        prompt = prompt_template.format(impersonation_template=impersonation_template, answer_option_template=answer_option_template, statement=political_statement, rag_template=rag_template)
     
     elif prompt_manipulation == "Most similar RAG (indirect steering with related context)":
         impersonation_template = ""
@@ -92,8 +97,7 @@ def predict(
         contexts = [context for context in retrieved_context['documents']]
         rag_template = f"\nHier sind Kontextinformationen:\n" + "\n".join([f"{context}" for context in contexts[0]])
         
-        prompt = prompt_template.format(impersonation_template=impersonation_template, answer_option_template=answer_option_template, statement=political_statement[3:], rag_template=rag_template)
-        print(prompt)
+        prompt = prompt_template.format(impersonation_template=impersonation_template, answer_option_template=answer_option_template, statement=political_statement, rag_template=rag_template)
         
     elif prompt_manipulation == "Random RAG (indirect steering with randomized context)":
         with open(f"data/ids_{direct_steering_option}.json", "r") as file:
@@ -109,15 +113,13 @@ def predict(
         contexts = [context for context in retrieved_context['documents']]
         rag_template = f"\nHier sind Kontextinformationen:\n" + "\n".join([f"{context}" for context in contexts])
         
-        prompt = prompt_template.format(impersonation_template=impersonation_template, answer_option_template=answer_option_template, statement=political_statement[3:], rag_template=rag_template)
-        print(prompt)
+        prompt = prompt_template.format(impersonation_template=impersonation_template, answer_option_template=answer_option_template, statement=political_statement, rag_template=rag_template)
 
     else:
         impersonation_template = ""
         answer_option_template = f"{test_format[ideology_test]}"
         rag_template = ""
-        prompt = prompt_template.format(impersonation_template=impersonation_template, answer_option_template=answer_option_template, statement=political_statement[3:], rag_template=rag_template)
-        print(prompt)
+        prompt = prompt_template.format(impersonation_template=impersonation_template, answer_option_template=answer_option_template, statement=political_statement, rag_template=rag_template)
 
     responses = []
     for model in [model1, model2]: