Aspect-Based Sentiment Analysis Ulasan Pengguna Aplikasi Polri Presisi
Abstract
Abstrak :
Pemanfaatan aplikasi mobile dalam layanan publik memerlukan evaluasi kualitas layanan berbasis persepsi pengguna. Penelitian terdahulu umumnya melakukan klasifikasi sentiment secar umum, atau memisahkan identifikasi aspek dan klasifikasi sentiment tanpa menghubungkannya dengan metrik prioritas perbaikan yang terukur. Penelitian ini bertujuan menjawab celah tersebut dengan mengusulkan pipeline Aspect-Based Sentiment Analysis (ABSA) dengan melakukan integrasi antara Latent Dirichlet Allocation (LDA), IndoBERT, Net Reputation Score (NRS) dan Long Short-Term Memory (LSTM) pada Aplikasi Polri Presisi. Data berupa 2640 ulasan pengguna versi 2.0.27 dikumpulkan dari Google Play Store dan menjadi 2458 ulasan setelah text preprocessing dengan pustaka Sastrawi dan kamus normalisasi slang Bahasa IndonesiaI. Identifikasi aspek menggunakan LDA dengan koherensi C_v menghasilkan empat aspek utama, yaitu Pengurusan SKCK, Efisiensi Pendaftaran, Isu Teknis, dan Kepuasan Layanan. IndoBERT berfungsi sebagai penganotasi sentiment otomatis yang divalidasi dengan rating pengguna. Prioritas perbaikan ditentukan menggunakan NRS. Aspek Isu Teknik memperoleh NRS -19,84% dan menjadi prioritas utama, diikuti Efisiensi Pendaftaran (51,51%), Pengurusan SKCK (75,78%) dan Kepuasan Layanan (77,57%). Klasifikasi topik dan sentimen dievaluasi melalui LSTM dengan pembagian data terstratifikasi 80:20, menghasilkan akurasi 85,07 untuk topik dan 86,91% untuk sentiment. Kontribusi penelitian ini meliputi pipeline ABSA terpadu yang mengintegrasi pemodelan topik, klasifikasi sentimen berbasis transformer, evaluasi deep learning, dan metrik prioritas NRS, penerapan pada domain layanan publik kepolisian digital yang masih terbatas kajiannya, serta pemberian dasar kuantitatif bagi pengelola aplikasi pemerintah dalam mengambil Keputusan perbaikan.
=============================================
Abstract :
The use of mobile applications in public services requires evaluation of service quality based on user perception. Previous studies generally perform sentiment classification in a general manner, or separate aspects and sentiment classification without linking them to the measured improvement priority metrics. This study aims to address this gap by proposing an Aspect-Based Sentiment Analysis (ABSA) pipeline by integrating Latent Dirichlet Allocation (LDA), IndoBERT, Net Reputation Score (NRS), and Long Short-Term Memory (LSTM) in the Presisi Police Application. Data in the form of 2,640 user reviews of version 2.0.27 were collected from the Google Play Store and reduced to 2,458 reviews after text preprocessing with the Sastrawi library and the Indonesian slang dictionary. Aspect identification using LDA with C_v coherence resulted in four main aspects: SKCK Management, Registration Efficiency, Technical Issues, and Service Satisfaction. IndoBERT functions as an automatic sentiment annotator validated with user ratings. Improvement priorities are determined using NRS. The Technical Issues aspect obtained an NRS of -19.84% and became the top priority, followed by Registration Efficiency (51.51%), SKCK Management (75.78%) and Service Satisfaction (77.57%). Topic and sentiment classification were evaluated through LSTM with stratified data splitting 80:20, resulting in an accuracy of 85.07 for topics and 86.91% for sentiment. The contribution of this research includes an integrated ABSA pipeline that integrates topic modeling, transformer-based sentiment classification, deep learning evaluation, and NRS priority metrics, application to the digital police public service domain which is still limited in its studies and providing a quantitative basis for government application managers in making improvement decisions.
Keywords
Full Text:
PDFReferences
[1] R. Fatmasari, W. Gata, N. K. Wardhani, K. Prayogi, and M. B. Husna, “Klasifikasi sentimen terhadap kualitas aplikasi bahan ajar digital akademik Universitas Terbuka di Google Play,” Jurnal SAINTEKOM, vol. 14, no. 1, pp. 48–60, 2024.
[2] D. Fristtikasari, S. Alam, and I. Kurniawan, “Analisis sentimen pengguna aplikasi Kitalulus pada ulasan Google Play Store menggunakan metode Naïve Bayes,” Jurnal Teknologi Informasi dan Komputer, vol. 10, no. 2, pp. 458–473, 2024.
[3] I. K. S. A. Rizaldi and S. Alam, “Analisis sentimen pengguna aplikasi JMO (Jamsostek Mobile) pada Google Play Store menggunakan metode Naïve Bayes,” STORAGE: Jurnal Ilmiah Teknik dan Ilmu Komputer, vol. 3, no. 1, pp. 45–58, 2024.
[4] S. Poria, E. Cambria, D. Hazarika, and P. Vij, “Aspect-based sentiment analysis: A survey,” IEEE Computational Intelligence Magazine, vol. 11, no. 2, pp. 21–35, May 2016.
[5] D. M. Blei, A. Y. Ng, and M. I. Jordan, “Latent Dirichlet allocation,” Journal of Machine Learning Research, vol. 3, pp. 993–1022, Jan. 2003.
[6] B. Wilie, K. Vincentio, G. I. Winata, S. Cahyawijaya, X. Li, Z. Y. Lim, S. Soleman, R. Mahendra, P. Fung, S. Bahar, and A. Purwarianti, “IndoNLU: Benchmark and resources for evaluating Indonesian natural language understanding,” in Proc. 1st Conf. Asia-Pacific Chapter Assoc. Comput. Linguistics and 10th Int. Joint Conf. Nat. Lang. Process. (AACL-IJCNLP), 2020, pp. 843–857.
[7] S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Computation, vol. 9, no. 8, pp. 1735–1780, Nov. 1997.
[8] M. Hu and B. Liu, “Mining and summarizing customer reviews,” in Proc. 10th ACM SIGKDD Int. Conf. Knowledge Discovery and Data Mining (KDD), Seattle, WA, USA, 2004, pp. 168–177.
[9] R. Hidayat and W. Gata, “Pemanfaatan IndoBERT dan Long Short-Term Memory (LSTM) pada analisis sentimen ulasan aplikasi Depok Single Window,” Information System for Educators and Professionals, vol. 10, no. 1, pp. 13–24, 2025.
[10] A. Hammami, “Aspect-based sentiment analysis ulasan pengguna aplikasi Identitas Kependudukan Digital Indonesia,” Jurnal Statistika dan Aplikasinya, vol. 8, no. 2, pp. 121–134, 2024.
[11] H. A. Tyas, R. Sarno, and B. S. Rintyarna, “Analisis perbandingan metode klasifikasi sentimen berita saham: Pendekatan machine learning, deep learning, transfer learning, dan graf,” Register: Jurnal Ilmiah Teknologi Sistem Informasi, vol. 10, no. 2, pp. 145–158, 2024.
[12] J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training of deep bidirectional transformers for language understanding,” in Proc. 2019 Conf. North Am. Chapter Assoc. Comput. Linguistics: Human Language Technologies, Minneapolis, MN, USA, 2019, pp. 4171–4186.
[13] H. Jelodar, Y. Wang, R. Orji, and X. Zeng, “Latent Dirichlet allocation (LDA) for topic modeling: A review,” Neurocomputing, vol. 358, pp. 290–305, Sep. 2019.
[14] D. Ma, S. Li, Y. Wang, and B. Li, “Interactive attention networks for aspect-level sentiment classification,” in Proc. 26th Int. Joint Conf. Artificial Intelligence (IJCAI), Melbourne, Australia, 2017, pp. 4068–4074.
[15] A. Khan, B. Baharudin, and A. Khan, “Text preprocessing and feature engineering for sentiment analysis: A survey,” International Journal of Advanced Computer Science and Applications, vol. 11, no. 7, pp. 160–170, 2020.
[16] M. Schuster and K. K. Paliwal, “Bidirectional recurrent neural networks,” IEEE Transactions on Signal Processing, vol. 45, no. 11, pp. 2673–2681, Nov. 1997.
[17] F. Reichheld, “The one number you need to grow,” Harvard Business Review, vol. 81, no. 12, pp. 46–54, Dec. 2003.
[18] T. Mikolov, K. Chen, G. Corrado, and J. Dean, “Efficient estimation of word representations in vector space,” in Proc. Int. Conf. Learning Representations (ICLR) Workshop, Scottsdale, AZ, USA, 2013, pp. 1–12.
[19] Y. Goldberg, “A primer on neural network models for natural language processing,” Journal of Artificial Intelligence Research, vol. 57, pp. 345–420, 2016.
[20] T. Y. Lin, P. Goyal, R. Girshick, K. He, and P. Dollár, “Focal loss for dense object detection,” in Proc. IEEE Int. Conf. Computer Vision (ICCV), Venice, Italy, 2017, pp. 2980–2988.
[21] N. V. Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, “SMOTE: Synthetic minority over-sampling technique,” Journal of Artificial Intelligence Research, vol. 16, pp. 321–357, 2002.
Refbacks
- There are currently no refbacks.





