Perbandingan Metode Word Embedding Untuk Analisis Sentimen Pada Data Ulasan Marketplace
Marketplace is a platform for buying and selling goods online, one of whichis shopee. The platform provides a lot of short text data about reviews of various products being sold. Therefore, sentiment analysis is carried out for the classification of reviews by taking into account the factors in the sentiment object.In sentiment analysis, there is a more advanced method, namely using word embedding, word representation in vectors, many researchers have used this method in their research. Therefore, this study uses review data obtained from the shopee marketplace for sentiment analysis.In this study, data is classified using Long Short Term Memory (LSTM).Reviews that are classified will have 2 labels namely positive and negative. Thisstudy aims to determine the final accuracy and vocabulary generated by word embedding which is classified using LSTM in analyzing sentiment in Indonesian shopee reviews.Word embedding methods used are Word2Vec and Global Vector (Glove).This study uses a dataset of 10,000 to produce a vocabulary of 18004 words. From the dataset, 80% training data and 20% test data were distributed. The accuracy of the word embedding word2vec method is 83% and the word embedding Glove method gets 86% accuracy.