Optimization of Text Mining Detection of Tajweed Reading Laws Using the Yolov8 Method on the Qur'an

Abstract

The science of tajweed is a science that studies how to read the letters or readings in the Qur'an beautifully or well by the legal rules regulated therein. However, many people still do not pay attention to the legal rules of tajweed when reading the Qur'an, so it is not uncommon for them to make mistakes in pronunciation. From the legal rules of Tajweed reading, the slightest difference will change the meaning and intended meaning of the reading. So, paying attention to every rule of the law of reading Tajweed is very important. Therefore, considering the current technological advances, we plan a tajweed detection design using the YOLO algorithm optimized for the Qur'an. This study aims to determine and analyze the detection of text mining on tajweed reading. The method used in this study is the YOLO Algorithm method. This research uses 210 images of the Mushaf Al-Qur'an dataset, tested twice using Augmentation and Non-Augmentation to get optimal research results. The dataset underwent a training process of 138 images, or about 66%, and a validation process of 48 images, about 28%, and 24 images, or 11% of the total sample. Of the two tests using augmentation with no augmentation, augmentation testing produces the highest precision value with a value of 0.985 or 98.5% and the highest mAP50 with a value of 0995 or 99.5% for the Lafdzul Jalalah class group, with a total accuracy value of 92.94%. For testing without augmentation, the results show that the highest mAP50 value is the Lafdzul Jalalah class, with a value of 0.974 or 97.40% and an accuracy value of 91.37%. Based on optimization and comparison carried out for the accuracy value of research with augmentation of 92.94% and research conducted without augmentation is 91.37%. So, the study's results obtained an increased value of 1.57% by performing greyscale augmentation.