Spaces:
Sleeping
Sleeping
Download src/neural_parser.py from Akshat20051/Dnn_project: direct link, hf CLI and curl.
- Browser
- Download file 5.69 kB
-
https://huggingface.co/spaces/Akshat20051/Dnn_project/resolve/main/src/neural_parser.py
- Command line
-
hf download hf://spaces/Akshat20051/Dnn_project/src/neural_parser.py
-
curl -L -o neural_parser.py https://huggingface.co/spaces/Akshat20051/Dnn_project/resolve/main/src/neural_parser.py
5.69 kB
| import pdfplumber | |
| import pandas as pd | |
| import re | |
| import os | |
| from getpass import getpass | |
| from src.logger import logger | |
| # Try importing OCR dependencies | |
| try: | |
| import pytesseract | |
| from pdf2image import convert_from_path | |
| OCR_AVAILABLE = True | |
| except ImportError: | |
| OCR_AVAILABLE = False | |
| logger.warning("OCR Dependencies (pytesseract/pdf2image) not found. Scanned PDFs cannot be processed.") | |
| # Delayed import to avoid circular dependency | |
| def get_classifier(): | |
| from src.classifier import tag_transactions | |
| return tag_transactions | |
| def parse_with_neural_engine(pdf_path): | |
| logger.info(f"🧠 NEURAL ENGINE: Scanning {pdf_path}...") | |
| transactions = [] | |
| pdf = None | |
| try: | |
| # 1. ATTEMPT OPENING | |
| try: | |
| pdf = pdfplumber.open(pdf_path) | |
| except Exception as e: | |
| if "password" in str(e).lower() or "authenticate" in str(e).lower(): | |
| logger.warning("File is encrypted.") | |
| user_pass = getpass(" [PROMPT] Enter PDF Password: ") | |
| pdf = pdfplumber.open(pdf_path, password=user_pass) | |
| else: | |
| logger.error(f"PDF Initialization Failed: {str(e)}") | |
| return None | |
| # 2. VALIDATE PAGES | |
| if not pdf.pages: | |
| logger.error("PDF contains no pages.") | |
| return None | |
| # 3. TEXT EXTRACTION (Hybrid: Text -> OCR) | |
| full_text = "" | |
| for i, page in enumerate(pdf.pages): | |
| text = page.extract_text() | |
| if text and len(text.strip()) > 50: | |
| full_text += text + "\n" | |
| else: | |
| logger.warning(f"Page {i+1} seems to be an image. Attempting OCR...") | |
| if OCR_AVAILABLE: | |
| try: | |
| # Convert specific page to image | |
| # Note: poppler_path might need to be configured on Windows, but on Linux usually auto-detected | |
| images = convert_from_path(pdf_path, first_page=i+1, last_page=i+1) | |
| for img in images: | |
| ocr_text = pytesseract.image_to_string(img) | |
| full_text += ocr_text + "\n" | |
| logger.info(f"OCR extracted {len(ocr_text)} chars from Page {i+1}") | |
| except Exception as ocr_e: | |
| logger.error(f"OCR Failed for Page {i+1}: {ocr_e}") | |
| logger.info("Ensure 'poppler-utils' and 'tesseract-ocr' are installed.") | |
| else: | |
| logger.error("OCR is not available (dependencies missing). Skipping page.") | |
| if not full_text.strip(): | |
| logger.error("No extractable text found. If this is a scanned photo, ensure OCR is set up.") | |
| return None | |
| lines = full_text.split('\n') | |
| logger.info(f"Analyzing {len(lines)} text lines...") | |
| # 4. PATTERN RECOGNITION (REGEX) | |
| # Enhanced to be more robust for OCR noise | |
| date_pattern = re.compile(r'\b(\d{1,2}[-/\s][A-Za-z]{3}[-/\s]\d{2,4})\b') | |
| money_pattern = re.compile(r'[\d,]+\.\d{2}') # Matches 12,345.00 | |
| current_txn = {} | |
| for line in lines: | |
| # Cleanup noise | |
| cleaned_line = line.strip() | |
| if not cleaned_line: continue | |
| match = date_pattern.search(cleaned_line) | |
| if match: | |
| if current_txn: transactions.append(current_txn) | |
| date_str = match.group(0) | |
| remaining = cleaned_line.replace(date_str, "") | |
| money = money_pattern.findall(remaining) | |
| val = 0.0 | |
| if money: | |
| # Heuristic: Usually the last money amount is the balance, second last is transaction amount | |
| # But in statements, sometimes it's Credit | Debit | Balance. | |
| # This logic is simplistic but works for many standard formats. | |
| # OCR might introduce commas or dots wrongly. | |
| amt_str = money[-1] if len(money) == 1 else money[-2] if len(money) > 1 else "0" | |
| try: | |
| val = float(amt_str.replace(',', '')) | |
| except: | |
| val = 0.0 | |
| # Determine Dr/Cr | |
| is_deposit = any(x in cleaned_line.lower() for x in ["cr", "credit", "deposit"]) | |
| current_txn = { | |
| "Date": date_str, | |
| "Description": re.sub(r'\s+', ' ', remaining).strip(), | |
| "Withdrawal": val if not is_deposit else 0.0, | |
| "Deposit": val if is_deposit else 0.0 | |
| } | |
| elif current_txn: | |
| current_txn['Description'] += " " + cleaned_line | |
| if current_txn: transactions.append(current_txn) | |
| df = pd.DataFrame(transactions) | |
| if not df.empty: | |
| logger.info(f"SUCCESS: Reconstructed {len(df)} transactions.") | |
| # Real-time categorization | |
| try: | |
| tagger = get_classifier() | |
| df = tagger(df) | |
| logger.info("🧠 NEURAL ENGINE: Auto-categorized transactions via AI.") | |
| except Exception as e: | |
| logger.warning(f"AI Categorization during parsing failed: {e}") | |
| return df | |
| else: | |
| logger.error("Logic failed to find dates in text.") | |
| return None | |
| except Exception as e: | |
| logger.critical(f"System Crash in Parser: {str(e)}") | |
| return None | |
| finally: | |
| if pdf: | |
| pdf.close() | |