Implementasi Optical Character Recognition dan Algoritma Pencocokan pada Validasi Dokumen Penyumpahan Advokat

The documents required for the swearing-in of advocates at Pengadilan Tinggi Padang are still checked manually, making the process time-consuming and prone to recording errors. This study aimed to develop and evaluate a document validation module that integrates Tesseract-based optical character recognition (OCR) with matching algorithms appropriate to each type of data element. The system was developed using the waterfall model as a Laravel-based web application connected to a Python-based OCR microservice and equipped with seven image preprocessing variants. Matching was performed using the Levenshtein distance algorithm for free-text elements and exact matching for elements with standardized formats. Testing involved 74 samples from 10 document types, covering 101 data elements. The test results showed that the overall extraction accuracy reached 89.1%, with the lowest accuracy recorded for Kartu Tanda Penduduk (KTP) documents, at 66.7%. At the matching stage, both algorithms achieved precision, recall, and accuracy values of 100% each. The discrepancies identified in seven data elements resulted from extraction errors rather than the matching logic. These findings indicate that OCR extraction quality is an important aspect of supporting accurate document validation. The developed module serves as a tool to assist in screening the documents required for the swearing-in of advocates, while administrative decisions remain within the authority of legal registry staff.