Digital Data Classification and Extraction for Records Management of PAPS and PACS Documents

Abstract

The Department of Social Welfare and Development – Adoption Resource and Referral Unit is in need of emerging technologies to store records of the Alternate Parental Care program. A technology that would digitize data, case visualize and notify managers and Server-Based digital storage. To hasten the storing process, the system does the classification and extraction of data from a PDF document using Tesseract Optical Character Recognition (OCR). OCR extracts the text from an image by converting the whole image into binary pixels and from binary pixels and compare the recognized sets to its library to predict the text and set of words. It uses Keras to train the model with tensorflow on the backend to classify data. Keras works by defining a sequence of layers in a network by creating a sequential class and adding new layers. This will be done by creating an array of layers and pasting it to the constructor of the sequential model. The researchers produced 4 four documents with ten pages of the same content in different resolutions. Based on the result gathered, 75 DPI resolution has an accuracy percentage of 94.8548778, for 150 DPI is 99.23619895, while 98.91023618 for 300 DPI and 98.7416335 for word application generated document. For 75 DPI, the error contains almost English words, while in 150 DPI and 300 DPI the error usually came from punctuation marks and in word application generated document, the error usually came from a capitalized letter.