Deteksi Emosi pada Sentimen Ulasan Pengguna Shopee Paylater Menggunakan Emotion Lexicon (EMOLEX) dan Long Short Term Memory (LSTM)

Abstract

User reviews are real feedback that can give new users an insight before they use a service. However, reading a lot of reviews can take a long time and the reviews you read can be biased. Shopee PayLater is one of the payment methods available on the Shopee app, allowing users to purchase items with a "buy now, pay later" system or installments according to their preferences during the transaction. This study aims to determine the emotions in user sentiment reviews of Shopee PayLater using Emotion Lexicon and Long Short Term Memory (LSTM). Sentiments are categorized into positive, negative, and neutral. The study further breaks down these sentiments into eight basic emotions: anger, anticipation, disgust, fear, joy, sadness, surprise, and trust, based on the Emotion Lexicon. The pre-processing stage in this study includes cleaning, case folding, normalization, stopword removal, and stemming. The results of this study show that the LSTM model testing achieved an accuracy of 0.88 with a batch size of 32, 8 units, and 22 epochs. In positive sentiment, the emotion joy was the highest 27.79%, followed by trust 27.52%, anticipation 23.16%, surprise 11.72%, fear 5.18%, sadness 2.72%, anger 1.36%, disgust 0.54%. In negative sentiment, the highest emotion was fear 23.63%, followed by sadness 21.98%, anger 14.84%, trust 10.99%, disgust 9.34%, surprise 8.24%, anticipation 6.87%, joy 4.12%. In neutral sentiment, trust was the highest emotion 23.71%, followed by sadness 14.95%, anticipation 13.92%, joy 11.86%, fear 11.34%, anger 9.79%, surprise 8.76%, disgust 5.67%.