An Improved Algorithm and Accuracy Analysis Testing Cases of Latin-to-Balinese Script Transliteration Method based on Bali Simbar Dwijendra Font
As part of the Balinese culture, the Balinese script has rarely been used today. Its Latin-to-Balinese script transliteration knowledge is also affected by that condition. This study aims to preserve that knowledge through technological approaches. By collaboration between informatics and linguistics discipline, this research improves the Latin-to-Balinese script transliteration method from the previous research. This method can be used as a core for the web/mobile-based transliteration learning application. The improvements that were taken in this research are: 1) the utilization of Bali Simbar Dwijendra font rather than the previous Bali Simbar-B font; and 2) additional accuracy analysis testing cases to the previous accuracy analysis testing cases from The Balinese Alphabet writing rules and examples document. As main contribution in this research area, this method was considered as the first Latin-to-Balinese script transliteration method that utilizing Bali Simbar Dwijendra font. This font is the enhancement from the previous Bali Simbar-B font related to its size which are bigger and thicker for better visual at the same font size. With certain keyboard layout application, it also makes easier and more natural to type it in programs and across the web. Additional 47 testing cases to the previous 151 testing cases related to the sound killers (gantungan) that has no example at The Balinese Alphabet writing rules and examples document. Those examples were taken from another credible source, i.e. the Balinese script dictionary from the Bali Province government. Through the experiment, this method has accuracy up to 98% (195 of 198) testing cases that has passed through all additional 47 cases, and 148 cases from the previous 151 cases. Related to the previous 151 cases, this improved method maintained same accuracy rate to the previous method. Execution time for relatively more complex sentence case type gave relatively real time result at 10.99ms on average on certain hardware and Operating System platform. In the future work, this method accuracy could be improved significantly by: 1) enriching the repository consists of words belong to special words from the Balinese script dictionary; and 2) implementing semantic relation transliteration.