COMPARISON OF THAI WORD SEGMENTATION ALGORITHMS FOR SENTIMENT ANALYSIS OF TIKTOK APPLICATION USERS Research Article

Abstract

The purposes of the research were to 1) compare the performance of Thai word segmentation algorithms in the PyThaiNLP library: Maximum Matching (MM), New Maximum Matching (NewMM), and Longest Matching (Longest) , in combination with stopword removal using PyThaiNLP stopword and custom stopword defined by the researcher; and 2) evaluate the performance of models Naïve Bayes (NB), Logistic Regression (LR), Support Vector Machine (SVM), and Long Short-Term Memory (LSTM) for classifying user sentiments in TikTok user reviews. The study used accuracy, precision, recall, F1-score, and training time as performance indicators. The dataset consisted of 5,519 customer reviews from the Google Play Store. The results revealed that 1) the NewMM segmentation algorithm achieved the best results, with an accuracy of 83.52% and an F1-score of 83.06%. 2) The Naïve Bayes achieved the best performance, with an accuracy of 83.73% and an F1-score of 83.43%. 3) Not removing stopwords always led to the best performance. These findings demonstrate the importance of selecting appropriate word segmentation algorithms and methods of stopword treatment to maximize the performance of sentiment analysis for Thai text.