Mutiara Oktrilisa
Universitas Adzkia

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

Aspect-Based Sentiment Mapping of Google Reviews for Digital Reputation Insights of Private Higher Education Institutions in Padang City Aldo Eko Syaputra; Sopi Sapriadi; Wahyu Prima; Yofhanda Septi Eirlangga; Mutiara Oktrilisa
Journal of Information System and Informatics Vol 8 No 5 (2026): October
Publisher : Asosiasi Doktor Sistem Informasi Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.63158/journalisi.v8i5.1830

Abstract

Online reviews provide public feedback for higher education institutions, but overall star ratings cannot explain which institutional aspects shape public perception. This study aims to provide partial and aspect-specific digital reputation insights for five private higher education institutions in Padang City through aspect-based sentiment mapping of textual Google Reviews. Using Outscraper, 1,159 Google Review records were collected. After excluding 794 star-only reviews, 365 textual reviews were retained and transformed into 484 aspect-level records across five aspects: academic and lecturers, campus facilities, administrative services, environment and location, and general/others. The dataset was highly imbalanced, with 443 positive records (91.53%), 35 negative records (7.23%), and 6 neutral records (1.24%). Sentiment classification was evaluated using TF-IDF with Logistic Regression, TF-IDF with SVM, and IndoBERT through 5-fold stratified cross-validation. IndoBERT achieved the strongest performance, with an accuracy of 0.973, balanced accuracy of 0.710, macro F1-score of 0.760, and weighted F1-score of 0.971. The findings show that aspect-based sentiment mapping can support Google Review-based public perception monitoring by producing a university–aspect–sentiment map for the selected institutions. These findings are limited to textual Google Reviews from five selected institutions and should be validated using larger and more balanced datasets.