Aspect-Based Sentiment Analysis for Mobile App Review Using Convolutional Neural Network (CNN) and Word2Vec

The proliferation of mobile applications in today’s digital environment has revolutionized the way people interact with technology, their experiences are often reflected in reviews, providing a rich source of data for analysis. Therefore, it is important to analyze the sentiment of user reviews. However, sentiment analysis can only determine whether a review tends to be positive or negative without understanding the sentiment-related aspects of the review. Therefore, further analysis is needed to extract the aspects and sentiments expressed in the reviews. Aspect-Based Sentiment Analysis (ABSA) focuses on identifying aspects of entities in text and related sentiments and requires expanding features capable of representing relationships between words. Word2Vec, an alternative feature algorithm, takes a more nuanced approach by capturing the semantic relationships between words in user reviews. This research proposes a new method that combines Convolutional Neural Network (CNN) and Word2Vec to perform ABSA. The proposed method is used to analyze user reviews of the “MyPertamina” mobile application because the application was developed recently and is prone to instability and errors, causing commotion and concern among users. Therefore, this provides a suitable opportunity to utilize user reviews on the application to evaluate the proposed algorithm for performing ABSA. The sentiments considered in this study were only positive and negative. Meanwhile, the aspects considered are based on general user complaints which in this research are categorized as bugs, subsidies, and payments. Aspect Classification Modeling obtained an accuracy value of 71%. The results of ABSA show varying performance in various aspects—namely Bugs, Subsidies, and Payments, namely 88.3%, 73%, and 66.8%.