Topic-level Sentiment Analysis for User Reviews in Gasoline Subsidy Application

Darin Wijaya, Hendri Murfi, Gianinna Ardaneswari

Research output: Contribution to journalConference articlepeer-review

Abstract

As a State-Owned Enterprise in the energy sector, Pertamina must ensure the targeted distribution of subsidized gasoline and prevent misuse. In this effort, starting from July 1, 2022, Pertamina initiated a program called "Subsidi Tepat (appropriate subsidies)". Registration for the program is possible through the MyPertamina app, which is downloadable from the Play Store. By early March 2023, MyPertamina had been downloaded over 10 million times on the Play Store. However, its rating stands at only 2,9/5. Given the significant downloads and low ratings, user reviews require analysis to ensure the performance of MyPertamina. This research employs the topic-level sentiment analysis using the BERT-EFCM model to predict the topics and the BERT-NN model to assess sentiments expressed on each discussed topic. The study reveals three main topics regarding MyPertamina: the app's use for fuel at gas stations, registration and services associated with the application, and user evaluations. Most users express negative sentiments, with sentiment ratios of 84% negative and 16% positive for the first topic, 85% negative and 15% positive for the second, and 80% negative and 20% positive for the third.

Original languageEnglish
Pages (from-to)221-224
Number of pages4
JournalProceedings - Swiss Conference on Data Science, SDS
Issue number2024
DOIs
Publication statusPublished - 2024
Event11th IEEE Swiss Conference on Data Science, SDS 2024 - Zurich, Switzerland
Duration: 30 May 202431 May 2024

Keywords

  • Sentiment analysis
  • topic detection
  • BERT
  • clustering
  • EFCM

Fingerprint

Dive into the research topics of 'Topic-level Sentiment Analysis for User Reviews in Gasoline Subsidy Application'. Together they form a unique fingerprint.

Cite this