N-Gram Stemming for Dialect of Api Lampung

Stemming is the technique of extracting base words from affixed words to improve recall by minimizing the variance of affixed words down to their base word forms. We used the n-gram stemming method, in which the affixed and base words are converted to n-gram form, and the degree of similarity between the n-gram of the affixed word and the n-gram of the base word is measured using the dice coefficient method; if the degree of similarity meets the specified threshold value, the base word is displayed in comparison to the affixed word. In this study, we created a stemmer approach that combines all affix variants in the Lampung language dialect of Api using the n-gram stemming method. At first, we looked for the best threshold value for the N-Gram Stemming approach, which we found to be <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">${0. 5}$</tex> for bigram and trigram. Based on <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">${5 0 0}$</tex> test words, the Bigram accuracy score is 0.94 for a threshold of 0.5 and 0.91 for a threshold of 0.55, but the Trigram accuracy value is <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">${0. 8 8}$</tex> for a threshold of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">${0. 5}$</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">${0. 7 9}$</tex> for a threshold of 0.55. According to the results from 200 independent test words, the Bigram accuracy value is <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">${0. 9 3}$</tex> for threshold <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">${0. 5}$</tex> and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">${0. 8 9}$</tex> for threshold 0.55, while the Trigram accuracy value is 0.87 for threshold 0.5 and 0.80 for threshold 0.55. In this study, we were focused on the morphology of each word and the validation stage with a dictionary.