debug_extraction.py

#!/usr/bin/env python3
"""
Debug the extraction issues step by step.
"""

import re

def clean_answer(answer):
    """Clean answer for comparison."""
    if not answer:
        return ""
    
    answer = str(answer).strip()
    print(f"  clean_answer input: '{answer}'")
    
    # Remove common prefixes
    answer = re.sub(r'^(The answer is|Answer:|Final answer:?)\s*', '', answer, flags=re.IGNORECASE)
    print(f"  after prefix removal: '{answer}'")
    
    # Remove dollar signs and other currency symbols
    answer = re.sub(r'[\$\s]+', '', answer)
    print(f"  after dollar removal: '{answer}'")
    
    # Remove boxed formatting (handle nested cases like $\boxed{$70,000}$)
    answer = re.sub(r'\\boxed\{([^}]+)\}', r'\1', answer)
    print(f"  after boxed removal: '{answer}'")
    
    # Remove brackets, parentheses
    answer = re.sub(r'^[\[\](){}]+|[\[\](){}]+$', '', answer)
    print(f"  after bracket removal: '{answer}'")
    
    # Remove trailing punctuation (periods, commas, etc.)
    answer = re.sub(r'[.,;:!?]+$', '', answer)
    print(f"  after punctuation removal: '{answer}'")
    
    # Remove commas from numbers (e.g., "70,000" -> "70000")
    answer = re.sub(r',', '', answer)
    print(f"  after comma removal: '{answer}'")
    
    # Try to extract just the number
    number_match = re.search(r'([0-9]+(?:\.[0-9]+)?)', answer)
    if number_match:
        num_str = number_match.group(1)
        print(f"  extracted number: '{num_str}'")
        # Convert to float and back to remove unnecessary decimals
        try:
            num = float(num_str)
            if num == int(num):
                result = str(int(num))
            else:
                result = str(num)
            print(f"  final result: '{result}'")
            return result
        except ValueError:
            print(f"  ValueError, returning: '{num_str}'")
            return num_str
    
    print(f"  no number found, returning: '{answer.strip()}'")
    return answer.strip()

def extract_answer_improved(text):
    """Improved answer extraction that looks for the final answer more carefully."""
    if not text or not text.strip():
        return ""
    
    print(f"Extracting from text: {text[:100]}...")
    
    # Split into lines for analysis
    lines = text.strip().split('\n')
    print(f"Split into {len(lines)} lines")
    
    # Strategy 1: Look for explicit answer patterns in the last few lines
    explicit_patterns = [
        r"(?:The )?final answer is:?\s*([^\n]+)",
        r"(?:The )?answer is:?\s*([^\n]+)",
        r"Answer:?\s*([^\n]+)",
        r"Therefore,?\s*(?:the )?answer is:?\s*([^\n]+)",
        r"Thus,?\s*(?:the )?answer is:?\s*([^\n]+)",
        r"Hence,?\s*(?:the )?answer is:?\s*([^\n]+)",
        r"So,?\s*(?:the )?answer is:?\s*([^\n]+)",
    ]
    
    print("Strategy 1: Looking for explicit answer patterns in last 5 lines")
    for i, line in enumerate(lines[-5:]):
        print(f"  Line {len(lines)-5+i}: '{line}'")
        for pattern in explicit_patterns:
            match = re.search(pattern, line, re.IGNORECASE)
            if match:
                answer = match.group(1).strip()
                print(f"    Found match with pattern '{pattern}': '{answer}'")
                cleaned = clean_answer(answer)
                if cleaned and is_valid_answer(cleaned):
                    print(f"    Valid answer: '{cleaned}'")
                    return cleaned
                else:
                    print(f"    Invalid answer: '{cleaned}'")
    
    # Strategy 2: Look for boxed answers (check all lines, not just last few)
    boxed_patterns = [
        r"\\boxed\{([^}]+)\}",
        r"\$\$([^$]+)\$\$",
        r"\$([^$]+)\$",
    ]
    
    print("Strategy 2: Looking for boxed answers in all lines")
    for i, line in enumerate(lines):
        print(f"  Line {i}: '{line}'")
        for pattern in boxed_patterns:
            match = re.search(pattern, line)
            if match:
                answer = match.group(1).strip()
                print(f"    Found boxed match with pattern '{pattern}': '{answer}'")
                cleaned = clean_answer(answer)
                if cleaned and is_valid_answer(cleaned):
                    print(f"    Valid boxed answer: '{cleaned}'")
                    return cleaned
                else:
                    print(f"    Invalid boxed answer: '{cleaned}'")
    
    print("No valid answer found")
    return ""

def is_valid_answer(answer):
    """Check if an answer is valid."""
    if not answer or answer.strip() == "":
        return False
    
    try:
        num = float(answer)
        # Updated range to handle larger numbers like 70,000
        return -100000 <= num <= 1000000
    except ValueError:
        return False

# Test the problematic case
test_text = "The final answer is: $\\boxed{$70,000}$"
print("Testing problematic case:")
print(f"Input: {test_text}")
result = extract_answer_improved(test_text)
print(f"Result: '{result}'")