Accuracy Measurement on Indonesian Non-formal Affixed Word Stemming With Levenhstein

Abstract

Algorithm stemming of Nazief and Andriani has been developed in terms of speed and accuracy. One of the developments is Non-Formal Affix Algorithm. This algorithm increases the accuracy of stemming in non-formal affixed words. In its development, Indonesian language is used in two forms: standard and non-standard. Non-formal language is commonly used in non-formal situation such as in daily conversation, post, or comment on social media (Facebook, Twitter, Instagram, etc.). Non-formal Affix Stemming Algorithm is used to get the basic word of a non-formal word. However, this algorithm has limitation in non-formal affixed word stemming. Thus, this research aims to focus on the modification of non-formal affix algorithm to increase accuracy in non-formal affixed words stemming. The result of this research shows that modification of the combination of Levenshtein Distance and frequently used words dictionary has accuracy level of 88.33% with precision of 1 and error rate of 0.2. Meanwhile, Non-formal Affix algorithm has accuracy level of 73.3% at stemming 60 non-formal affixed words with precision of 1 and error rate of 0.26. Therefore, it can be concluded that Lenshtein Distance approach can increase the accuracy of Non-formal Affix algorithm in non-formal affixed words stemming.