Lontara Script

Based on the 'La Galigo NBG Boeg 188' manuscript housed in the Leiden University Library. https://digitalcollections.universiteitleiden.nl/lagaligo This language model has been developed as part of the Lontara Digital project to make historical documents written in the Lontara script more accessible. The model is based on a ground truth dataset derived from the La Galigo NBG Boeg 188 archive, housed at the Leiden University Library. The La Galigo NBG Boeg 188 archive is the largest continuous section of the La Galigo epic, consisting of 12 volumes and covering the initial third of this monumental Buginese literary work. The epic, written entirely in the Lontara script, is one of the most voluminous literary works in the world. Due to its immense length, no complete manuscript exists. Recognized for its cultural significance, the La Galigo epic was inscribed in UNESCO’s Memory of the World register in 2011. The ground truth dataset comprises 25 pages of handwritten Lontara script, transcribed manually into approximately 14,000 abugida characters, equivalent to 30,000 Latin characters. The model achieves a Character Error Rate (CER) of 4.72%, meeting the minimum threshold for publication. This model serves as the first attempt to create a transcription tool for the Lontara script and is an important milestone for historical and linguistic research, especially for historians focusing on Eastern Indonesian history. We hope this model will benefit researchers, linguists, and enthusiasts working with the Lontara script and the rich cultural heritage of South Sulawesi, Indonesia. Acknowledgment: This project was supported by the Leiden University Center for Digital Humanities Small Grant 2024.