Indonesian Public Service App Review Dataset: Type and Aspect Classification (IKD, Mobile JKN, MyPertamina)

This dataset contains 6,000 user reviews written in Indonesian, collected from the Google Play Store for three Indonesian public service applications: Identitas Kependudukan Digital (IKD), Mobile JKN, and MyPertamina, representing the population administration, healthcare, and energy subsidy sectors respectively. Reviews were scraped using the google-play-scraper library, cleaned of duplicates, empty or emoji-only content, and spam/promotional text, then selected via stratified random sampling (2,000 reviews per application) to ensure balanced representation across apps. Each review is annotated with two independent labels: (1) Review Type : Bug Report, Feature Request, or Other; and (2) Aspect/Module : one of six categories: Login/OTP/Verification, Update Data, Transaction, General Performance, Core Service, or Other. Labels were generated using a generative AI model as the primary annotator and validated for reliability against two human annotators on a 300-review subset (5%) using Fleiss' Kappa (κ = 0.82 for Review Type, κ = 0.82 for Aspect/Module its "almost perfect" agreement per Landis & Koch). This dataset supports research on automated app review classification, software maintenance prioritization, and digital public service quality monitoring in the Indonesian-language context.