Analysis of Google Play Store's Sentiment Review on Waqf Digital Platform Using Fasttext Embedding

Waqf has an important role in the development and increase in welfare. In addition to reducing dependence on funds from the Indonesian government, waqf has also had a significant impact on reviving the economy, especially since the outbreak of the COVID-19 virus. The rapid advancement of technology has also transformed waqf, one of which is that people can donate waqf money online through several digital applications. But so far, several advantages and disadvantages are felt by application users. To make it easier to get information based on user experience, we propose to develop a model that can classify sentiments into positive, negative, and neutral automatically. Text classification using word embedding is the basis for getting the best performance results. Bag of Word (BOW) is a word embedding model that is often used, but this model is considered not optimal because it has disadvantages such as dependence on certain languages. Therefore, we suggest the fastText model minimizes dependency on pre-processing words and use 2 classification methods, namely SVM and KNN. This study aims to compare the performance results using the fastText model with conventional models that are often used, namely Bag of Word (BOW) and Term Frequency – Inverse Document Frequency (TF-IDF) to find the best accuracy value produced. Based on this research, it can be interpreted that in general, the fastText model can produce better performance than BOW and TF-IDF.