Sundanese Stemming using Syllable Pattern

Abstract

Stemming is a technique to return the word derivation to the root or base word. Stemming is widely used for data processing such as searching word indexes, translating, and information retrieval from a document in the database. In general, stemming uses a morphological pattern from a derived word to produce the original word or root word.  In the previous research, this technique faced over-stemming and under-stemming problems. In this study, the stemming process will be improved by the syllable pattern (canonical) based on the phonological rule in Sundanese. The stemming result for syllable patterns gets an accuracy of 89% and the execution of the test data resulted in 95% from all the basic words. This simple algorithm has the advantage of being able to adjust the position of the syllable pattern with the word to be stemmed. Due to some data shortage constraints (typo, loan-word, non-deterministic word with syllable pattern), we can improve to increase the accuracy such as adjusting words and adding reference dictionaries. In addition, this algorithm has a drawback that causes the execution to be over-stemming.