<![CDATA[Indonesia has many historical and cultural heritages in the form of ancient documents written in Arabic, Malay, or Jawi. There are six additional letters in Jawi, namely ca, nya, nga, pa, ga, and va to form Jawi’s vocabulary. This ancient Jawi document has suffered many quality degradations such as uneven lighting, varying contrast, blurred ink and writing, black spots, smudges, and several other glitches. Research on handwritten word recognition systems has been done extensively for Arabic scripts and various other writings. Most of the previous research focused on character level, word level, and document level. However, it still uses publicly available datasets such as IFN/ENIT, CVL dataset, IAM, and several other datasets with similar characteristics. Meanwhile, the Jawi dataset at the word level is still not available at this time. Therefore, this study examines the handwriting recognition system at the word level. The purpose of this study is to propose a new dataset (word Jawi dataset). It is hoped that this new dataset can become a more representative dataset. The process of creating a new dataset is carried out using a manual and semi�automatic approach. Furthermore, the document said Jawi will determine the Ground Truth (GT). This research produces a special dataset of words Jawi as many as 2,310 words.]]>
Copyrights © 2022
