Integrating Aspect-Based Sentiment Analysis with CSI–IPA for Telecommunications App Development

This study integrates Aspect-Based Sentiment Analysis with Customer Satisfaction Index (CSI) and Importance Performance Analysis (IPA) to determine priority development features in digital service applications in the telecommunications sector. A total of 10,000 MyIM3 user reviews were analyzed using sentiment analysis with Pseudo-labeling and Fine-Tuned IndoBERT, then from the results of the analysis, negative sentiment was mapped into several topics using LDA. The topic is used to compile question indicators based on the five dimensions of SERVQUAL. After the questionnaire data is declared valid and reliable, CSI and IPA analysis is carried out. A CSI value of 79.22% indicates that user satisfaction is in the "borderline" category, but several aspects still need to be improved, especially system updates (RS3), application attention to user needs (E2), and feature personalization (E3) which are in quadrant I (Concentrate Here). This hybrid approach offers novelty by demonstrating how ABSA and LDA can be systematically integrated with CSI and IPA to provide more comprehensive and user-oriented insights. The limitations of this study include focusing on negative sentiment data for feature exploration, as it is most relevant for identifying problems and opportunities for improvement and development of telecommunication digital services.