Traveling Salesman Problem Algorithm-Based Uniqueness Explainable Features of the Jawi-Arabic Manuscript for Digitalization
TL;DR
This study introduces an explainable feature representation approach based on the Traveling Salesman Problem (TSP) algorithm to enhance traceability for the machine knowledge using mathematical formula and computing algorithms due to the complexity of curvature in recognizing each character for manuscript digitization.
Abstract
Abstract From the perspective of Artificial Intelligence (AI) and machine learning feature in computer science, this article aims to describe the representation to capture uniqueness of the Jawi-Arabic character curvature features including angularities, elongation, flourishment, connection, and overlapping letters, which supports the digitization of historical manuscripts. Recognizing handwritten Jawi-Arabic calligraphic manuscripts remains a major challenge in digitization mainly due to complexity and uniqueness in the curvature. The isolated representation of character shape features is still challenging due to the varied and rich handwriting style in the historical manuscript when using machine learning methods. This study introduces an explainable feature representation approach based on the Traveling Salesman Problem (TSP) algorithm to enhance traceability for the machine knowledge using mathematical formula and computing algorithms due to the complexity of curvature in recognizing each character for manuscript digitization. Our contribution is an analysis of the unique characteristics of paleography curvature properties of the manuscript to provide knowledge of isolating each character using mathematical formulas and computer algorithms as a bridge to provide manuscript retrieval. This is a complement to machine learning approaches, as it provides more human expertise reasoning to leverage the knowledge base of historical manuscripts.
