Comparison of LDA and BERTopic in Identifying Public Issues in the MBG Program

Nur Hayati, Saikin Saikin, Hairul Fahmi

Abstract


The Free Nutritious Food Program (MBG) is a government policy that has generated various public responses and opinions on social media. The large amount of unstructured text data. This study aims to compare the performance of the Latent Dirichlet Allocation (LDA) and BERTopic methods in identifying public issues related to the MBG program on TikTok data. The dataset used amounted to 13,538 data obtained through a scraping process based on keywords related to MBG. The research stages include text preprocessing, bigram and trigram formation, text representation using TF-IDF and embedding, topic modeling, and evaluation using coherence score and topic diversity. The results showed that the LDA method produced better evaluation performance with a coherence score of 0.5098 and a topic diversity of 0.9000. Meanwhile, BERTopic produced a coherence score of 0.4133 and a topic diversity of 0.7667, but was able to produce topics that were more contextual and semantically representative. Based on these results, LDA is superior in terms of the stability and quality of word associations between topics, while BERTopic is more effective in understanding the context of issues in short and unstructured social media data.

Keywords


Topic Modeling; LDA; BERTopic; Social Media; Free Nutritious Food

Full Text:

PDF

References


M. Grootendorst, “BERTopic: Neural topic modeling with a class-based TF-IDF procedure,” 2020.

P. Koochemeshkian and N. Bouguila, “Integration of Neural Embeddings and Probabilistic Models in Topic Modeling Integration of Neural Embeddings and Probabilistic Models in Topic Modeling,” Appl. Artif. Intel., vol. 38, no. 1, 2024, doi: 10.1080/08839514.2024.2403904.

F. Bianchi, S. Terragni, and D. Hovy, “Pre-training is a Hot Topic : Contextualized Document Embeddings Improve Topic Coherence,” pp. 759–766, 2021.

AB Dieng and DM Blei, “Topic Modeling in Embedding Spaces,” 2017.

S. Koltcov, A. Surkov, V. Filippov, and V. Ignatenko, “Topic models with elements of neural networks : investigation of stability, coherence, and determining the optimal number of topics,” pp. 1–41, 2024, doi: 10.7717/peerj-cs.1758.

L. Yijia, “Comparison of LDA and BERTopic in News Topic Modeling : A Case Study of The New York Times' Reports on China,” vol. 7, no. March, pp. 47–51, 2024, doi: 10.55014/pij.v7i3.616.

MD Pratiwi, KD Tania, S. Informasi, and U. Sriwijaya, “Knowledge Discovery Through Topic Modeling on GoPartner User Reviews Using BERTopic, LDA, and NMF,” vol. 9, no. 1, pp. 1–7, 2025.

Egger, J. Yu, and J. Yu, “A Topic Modeling Comparison Between LDA , NMF , Top2Vec , and BERTopic to Demystify Twitter Posts Making Sense of Social Media Using,” vol. 7, no. May, pp. 1–16, 2022, doi: 10.3389/fsoc.2022.886498

Luiz and M. Owa, "Identification of Topics from Scientific Papers through Topic Modeling," pp. 541–548, 2021, doi: 10.4236/ojapps.2021.114038.

CB Pavithra and J. Savitha, “Topic Modeling for Evolving Textual Data Using LDA , HDP , NMF , BERTOPIC , and DTM With a Focus on Research Papers,” vol. 5, no. 2, 2024, doi: 10.37802/joti.v5i2.618.

P. Resnik, “Improving Neural Topic Models using Knowledge Distillation,” pp. 1752–1771, 2020.

A. Farea, S. Tripathi, G. Glazko, and F. Emmert-streib, “Engineering Applications of Artificial Intelligence Investigating the optimal number of topics by advanced text-mining techniques : Sustainable energy research,” Eng. Appl. Artif. Intel., vol. 136, no. PA, p. 108877, 2024, doi: 10.1016/j.engappai.2024.108877.

X. Wu, T. Nguyen, and AT Luu, “and challenges,” Artif. Intel. Rev., vol. 57, no. 2, pp. 1–30, 2024, doi: 10.1007/s10462-023-10661-7.

MR Haas and A. Heijn, “Experiments on Generalizability of BERTopic on Multi-Domain Short Text,” pp. 1–3, 2020.

MC Kenton, L. Kristina, and J. Devlin, “BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,” no. MLM, 1953.

E. Taheri and JL Junkins, “How Many Impulses Redux”.

M. Röder and A. Hinneburg, “Exploring the Space of Topic Coherence Measures”.

P. Aprilio, PS Nugraha, and H. Fahmi, "Hybrid Feature Combination of TF-IDF and BERT for Enhanced Information Retrieval Accuracy," vol. 08, no. 01, pp. 8–15, 2025.

DM Blei, AY Ng, and MI Jordan, “Latent Dirichlet Allocation,” vol. 3, pp. 993–1022, 2003.

I. Computer and T. Hofmann, “Probabilistic Latent Semantic Indexing,” pp. 50–57.

COO Optimization, “Topic Modeling as Multi-Objective Contrastive Optimization,” pp. 1–20, 2024.

21] PMSArdinata, AAJ Permana, and INSW Wijaya, “IDENTIFICATION AND NORMALIZATION OF SLANG TEXT WITH,” vol. 21, no. 1, 2024.

[22] RLNurdiansyah and KE Dewi, “KOMPUTA: Scientific Journal of Computers and Informatics THE EFFECT OF INFORMATION GAIN AND WORD NORMALIZATION ON ASPECT-BASED SENTIMENT ANALYSIS KOMPUTA: Scientific Journal of Computers and Informatics,” vol. 12, no. 2, 2023.

R. Li, F. González-pizarro, L. Xing, G. Murray, and G. Carenini, “Diversity-Aware Coherence Loss for Improving Neural Topic Models,” vol. 2, pp. 1710–1722, 2023.

R. Li, F. González-pizarro, L. Xing, G. Murray, and G. Carenini, “Diversity-Aware Coherence Loss for Improving Neural Topic Models,” vol. 2, pp. 1710–1722, 2023.




DOI: https://doi.org/10.31326/jisa.v9i1.2782

Refbacks

  • There are currently no refbacks.


Copyright (c) 2026 Nur Hayati, Saikin Saikin, Hairul Fahmi

Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

JOURNAL IDENTITY

Journal Name: JISA (Jurnal Informatika dan Sains)
e-ISSN: 2614-8404, p-ISSN: 2776-3234
Publisher: Program Studi Teknik Informatika Universitas Trilogi
Publication Schedule: June and December 
Language: English
APC: The Journal Charges Fees for Publishing 
Indexing:  EBSCO , DOAJ, Google Scholar, Arsip Relawan Jurnal Indonesia, Directory of Research Journals Indexing, Index Copernicus International, PKP Index, Science and Technology Index (SINTA, S4) , Garuda Index
OAI address: http://trilogi.ac.id/journal/ks/index.php/JISA/oai
Contact: jisa@trilogi.ac.id
Sponsored by: DOI – Digital Object Identifier Crossref, Universitas Trilogi

In Collaboration With: Indonesian Artificial Intelligent Ecosystem(IAIE), Relawan Jurnal Indonesia, Jurnal Teknologi dan Sistem Komputer (JTSiskom)

 

 

JISA (Jurnal Informatika dan Sains) is Published by Program Studi Teknik Informatika, Universitas Trilogi under Creative Commons Attribution-ShareAlike 4.0 International License.