PT. Gojek adalah perusahaan yang menyediakan layanan transportasi, pengantaran makanan, kurir, belanjaan, pembayaran, dan lainnya. Pengguna dapat menilai dan memberi ulasan aplikasi Gojek di Google Play Store. Namun hal ini seringkali kurang mencerminkan kualitas aplikasi secara akurat karena pengguna sering kali memberikan penelitian yang berbeda dari ulasan mereka. Penelitian ini menganalisis 1194 ulasan aplikasi Gojek di Google Play Store dari Oktober 2023 hingga Maret 2024 menggunakan metode Long Short Term Memory (LSTM) untuk mengevaluasi akurasi dan distribusi sentimen ulasan. Proses penelitian meliputi pengumpulan data melalui web scraping, text preprocessing, pelabelan menggunakan InSet Lexicon, dan representasi kata dengan Word2Vec. Model LSTM digunakan untuk klasifikasi sentimen dan dievaluasi dengan matriks akurasi, presisi, recall, dan f1-score. Hasil pelabelan menggunakan InSet Lexicon menunjukkan 43,13% ulasan negatif, 41,46% positif, dan 15,41% netral. Model LSTM mencapai akurasi 76,66%, presisi untuk ulasan positif 79,6%, negatif 83,6%, dan netral 27,0%. Recall untuk ulasan positif 84,3%, negatif 83,6%, dan netral 21,0%, sedangkan f1-score untuk ulasan positif 81,8%, negatif 83,6%, dan netral 23,6%.

PT Gojek is a company that provides transportation, food delivery, courier, grocery, payment, and other services. Users can rate and review the Gojek application on the Google Play Store. However, this often does not accurately reflect the quality of the application because users often provide different research from their reviews. This research analyzes 1194 Gojek application reviews on the Google Play Store from October 2023 to March 2024 using the Long Short Term Memory (LSTM) method to evaluate the accuracy and distribution of review sentiment. The research process includes data collection through web scraping, text preprocessing, labeling using InSet Lexicon, and word representation with Word2Vec. The LSTM model is used for sentiment classification and evaluated with accuracy, precision, recall, and f1-score metrics. Labeling results using InSet Lexicon show 43.13% negative reviews, 41.46% positive, and 15.41% neutral. The LSTM model achieved 76.66% accuracy, precision for 79.6% positive, 83.6% negative, and 27.0% neutral reviews. Recall for positive reviews was 84.3%, negative 83.6%, and neutral 21.0%, while f1-score for positive reviews was 81.8%, negative 83.6%, and neutral 23.6%.