Indonesia Analysis of Public Opinion on Disability Services Using Sentiment Analysis and Latent Dirichlet Allocation Topic Modeling Methods

Zahid Anugrah Muzaffar Rana, Inggil Tahta Haritza, M. Arif Fadhillah, Dwi Adi PurnamaNon-Scopus11 Juli 2025Afiliasi: ID

Abstrak

In 2023, the number of people with disabilities in Indonesia reached 8.5% of the total population. Despite this significant proportion, they continue to face barriers in accessing public services, education, healthcare, and employment. Therefore, there is an urgent need for public policy analysis supported by real-time insights into public perceptions. This study aims to analyze online data from Twitter to inform policy recommendations for enchancing disability services. IndoRoBERTa fine-tuning was applied to capture nuanced emotional polarity in Indonesian-language texts, while Latent Dirichlet Allocation (LDA) topic modeling was employed to identify the latent themes behind these opinions. The two methods were deliberately combined because sentiment analysis alone cannot uncover the substantive issues underlying public views, while topic modeling alone cannot show the polarity of those views together; they provide a more comprehensive analytical framework. The novelty of this research lies in integrating advanced sentiment analysis and topic modeling to support the formulation of disability policy in the Indonesian context, which remains underexplored. A total of 18,242 tweets were collected using four keywords: disability, disability services, disability facilities, and disability programs. The analysis revealed critical issues, including physical accessibility, bullying prevention, and educational programs to reduce early marriage. The proposed model achieved strong performance with 97.42% accuracy, and precision, recall, and F1-scores all exceeded 95%, surpassing previous studies with an accuracy of around 93%. These findings demonstrate that data mining of online public opinion can serve as a robust medium for formulating responsive public policies and enabling real-time monitoring of disability-related services.

Identifier

Scroll to Top