A method for solving scriptio continua in Javanese manuscript transliteration

Area of Science:

• Digital Humanities
• Computational Linguistics
• Information Science

Background:

• Javanese manuscripts are stored in Indonesian museums and libraries.
• Local scripts in these manuscripts are rarely used, necessitating tools for improved reading.
• Automatic manuscript image transliteration requires post-processing to merge syllables into words.

Purpose of the Study:

• To propose and evaluate backtracking-based methods for solving the scriptio continua problem in Javanese manuscript transliteration.
• To enhance the accuracy and efficiency of post-processing in manuscript image analysis.

Main Methods:

• Utilized a backtracking algorithm with depth-first search to identify and merge syllables into words.
• Implemented greedy and brute-force approaches for syllable concatenation.
• Tested methods on 17,687 syllables from the Hamong Tani book with a 49,801-word dictionary.

Main Results:

• Both greedy and brute-force methods achieved an accuracy of 81.64% in concatenating syllables.
• The greedy-based method demonstrated superior efficiency and performance compared to the brute-force method.
• The proposed methods provided satisfactory results in terms of computation and accuracy for Javanese manuscript transliteration.

Conclusions:

• Backtracking algorithms effectively address the scriptio continua challenge in Javanese manuscript post-processing.
• The greedy approach offers a more efficient and performant solution for syllable concatenation.
• The developed methods contribute to improving the accessibility and study of historical Javanese manuscripts.