Aspect-Based Sentiment Analysis of Transportation Electrification Opinions on YouTube Comment Data

Introduction/Main Objectives: This research aims to conduct an aspect-based sentiment analysis of transportation electrification opinions on YouTube comment data. Background Problems: It is difficult to summarize the sentiment of many YouTube user comments related to electric vehicles (EVs) based on their aspects; therefore, aspect-based sentiment analysis is needed to conduct further analysis. Novelty: This study identifies five aspects of EV and their sentiments at the same time. The aspects are usefulness, ease of use, comfort, cost, and incentive policies. One of this study’s methods is the transfer learning model. This model can be a solution to overcome the shortcomings of deep learning in classifying aspect-based sentiment classification on small datasets. Research Methods: The sentiment classification model used is a machine learning model, namely support vector machine (SVM) and transfer learning models from pre-trained IndoBERT and mBERT. Finding/Results: Based on the experimental results, transfer learning from the IndoBERT model achieved the best performance with accuracy and F1-Score of 89.17% and 52.66%, respectively. Furthermore, the best IndoBERT model was developed with input in the form of a combination of aspects and comment sentences. Experimental results show that there is an improvement in performance with accuracy and F1-Score of 90% and 60.70%, respectively.