Abstract This paper focused on improve results accuracy Alphabet text transliteration into Arabic text in Indonesian Malay. Alphabet text transliteration into Jawi in this research using reference books by Rosadi rule that takes into account the beheading of syllables for determination its Jawi text. The method used in this study to identify syllables is Fine State Automata (FSA) developed by Basuki, however the syllable identifiers can not distinguish beheading diphthong with the vowel sequence, this of course affects the transliteration accuracy. The method used in this study to improve the accuracy of transliteration that modify FSA for identifier Indonesian syllables that have been made by Basuki then added the ability to recognize patterns diphthong syllables and vowel series. Results obtained modifications that have been done syllable recognition accuracy improvement is 25% from the previous. The level of accuracy of the transliteration is 94.74% if calculated based on the results of beheading the right conditions, but when it counted the results obtained from incorrect beheading, the transliteration accuracy is 90%. Keywords: jawi, natural language processing, recognition syllables.
Copyrights © 2016