AI Document Processing

The Document Processing Challenge

Organizations deal with millions of unstructured documents:

Invoices and receipts

Contracts and legal documents

Medical records

Government forms

Financial reports

OCR with Tesseract + AI Enhancement

python
import pytesseract
from PIL import Image
import cv2
def extract_text_from_image(image_path: str) -> str:
    # Preprocess image for better OCR
    img = cv2.imread(image_path)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    denoised = cv2.fastNlMeansDenoising(gray, h=10)
    
    # Tesseract OCR
    pil_img = Image.fromarray(denoised)
    text = pytesseract.image_to_string(pil_img, config='--psm 6')
    return text
def enhance_with_ai(raw_text: str, document_type: str) -> dict:
    """Use AI to fix OCR errors and extract structure"""
    response = client.chat.completions.create(
        model="gpt-4o",
        messages=[{
            "role": "user",
            "content": f"""Clean up this {document_type} OCR text and extract:
            - All field names and values
            - Fix obvious OCR errors
            Return as structured JSON.
            
            Raw text: {raw_text}"""
        }],
        response_format={"type": "json_object"}
    )
    return json.loads(response.choices[0].message.content)

Azure Document Intelligence (Form Recognizer)

python
from azure.ai.formrecognizer import DocumentAnalysisClient
from azure.core.credentials import AzureKeyCredential
client = DocumentAnalysisClient(
    endpoint=AZURE_ENDPOINT,
    credential=AzureKeyCredential(AZURE_KEY)
)
with open("invoice.pdf", "rb") as f:
    poller = client.begin_analyze_document("prebuilt-invoice", f)
    result = poller.result()
invoice = result.documents[0]
print(f"Invoice Number: {invoice.fields['InvoiceId'].value}")
print(f"Total: {invoice.fields['InvoiceTotal'].value}")

LLM-Based Extraction with Schema

python
from pydantic import BaseModel
from typing import Optional, List
class InvoiceLineItem(BaseModel):
    description: str
    quantity: float
    unit_price: float
    total: float
class Invoice(BaseModel):
    invoice_number: str
    date: str
    vendor_name: str
    total_amount: float
    line_items: List[InvoiceLineItem]
    tax_amount: Optional[float] = None
def extract_invoice(text: str) -> Invoice:
    response = client.beta.chat.completions.parse(
        model="gpt-4o",
        messages=[
            {"role": "user", "content": f"Extract invoice data: {text}"}
        ],
        response_format=Invoice
    )
    return response.choices[0].message.parsed

Pipeline for Scale

python
from celery import Celery
app = Celery('document_processor')
@app.task
def process_document(doc_path: str, doc_type: str):
    raw_text = extract_text_from_image(doc_path)
    structured_data = extract_invoice(raw_text)
    save_to_database(structured_data)
    return {"status": "success", "data": structured_data.dict()}