Development of Cosine Dissimilarity-Based Article Recommendation System Using LLM-Generated Abstractive Summaries 


Vol. 51,  No. 2, pp. 390-399, Feb.  2026
10.7840/kics.2026.51.2.390


PDF Full-Text
  Abstract

With the rapid growth of online news and social media, modern society faces an overload of textual information. To address this issue, text summarization technologies are gaining increasing importance. This study proposes a cosine dissimilarity-based article recommendation system that utilizes abstractive summaries generated by a fine-tuned large-scale language model. Unlike traditional extractive summarization, which often suffers from fragmented information and lack of logical coherence, our system leverages KoBART, a pre-trained Korean language model, fine-tuned on the AIHub Korean summarization dataset. The model generates summaries that effectively reconstruct the core content of the original text while ensuring high readability and semantic clarity. By measuring cosine dissimilarity between generated summaries, the system recommends news articles with diverse perspectives, thereby mitigating information bias. The summarization performance was validated using ROUGE, SentenceTransformer, and SBERT evaluations, achieving over 0.67 and 0.77 in semantic similarity, respectively. Finally, a user-friendly web interface was implemented using Streamlit, allowing users to access article titles, original texts, summaries, and source links. The proposed system enhances both the diversity and accessibility of news consumption in a highly saturated information environment.

  Statistics
Cumulative Counts from November, 2022
Multiple requests among the same browser session are counted as one view. If you mouse over a chart, the values of data points will be shown.


  Related Articles
  Cite this article

[IEEE Style]

H. Choi and H. Lee, "Development of Cosine Dissimilarity-Based Article Recommendation System Using LLM-Generated Abstractive Summaries," The Journal of Korean Institute of Communications and Information Sciences, vol. 51, no. 2, pp. 390-399, 2026. DOI: 10.7840/kics.2026.51.2.390.

[ACM Style]

Hyeju Choi and Harim Lee. 2026. Development of Cosine Dissimilarity-Based Article Recommendation System Using LLM-Generated Abstractive Summaries. The Journal of Korean Institute of Communications and Information Sciences, 51, 2, (2026), 390-399. DOI: 10.7840/kics.2026.51.2.390.

[KICS Style]

Hyeju Choi and Harim Lee, "Development of Cosine Dissimilarity-Based Article Recommendation System Using LLM-Generated Abstractive Summaries," The Journal of Korean Institute of Communications and Information Sciences, vol. 51, no. 2, pp. 390-399, 2. 2026. (https://doi.org/10.7840/kics.2026.51.2.390)
Vol. 51, No. 2 Index

Search
(IN TITLE, AUTHOR, ABSTRACT,KEYWORDS)

Advanced Search ▼

POPULAR KEYWORDS
(TOP 10 KEYWORDS)

Recent Publications
(LAST 3 YEARS)






pISSN : 1226 - 4717
eISSN : 2287 - 3880