Claim Missing Document
Check
Articles

Found 1 Documents
Search

Graph-based text summarization using LexRank to uncover student perspectives on social media Sulfayanti F. Situju; Siti Aulia Rachmini; Nadia; Fitriyani Umar
Journal of Soft Computing Exploration Vol. 7 No. 4 (2026): December 2026
Publisher : SHM Publisher

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.52465/joscex.v7i4.161

Abstract

The large number of student comments challenged efficient manual analysis. This study addressed the problem by applying the LexRank algorithm to generate automatic summaries that captured the most important information. A summary of the student comment data was obtained through sentence segmentation, text preprocessing, TF-IDF weighting, cosine similarity calculation, graph construction, and sentence ranking using the LexRank algorithm. The summary results were evaluated using the ROUGE metric by comparing LexRank-generated summaries with those produced by generative AI. The test results showed that direct summarization of 500 comments simultaneously produced F1-scores of ROUGE-1 = 0.309, ROUGE-2 = 0.093, and ROUGE-L = 0.191. Additional experiments were conducted by dividing the data into five subsets, each containing 100 comments, and yielded the best F1-scores of ROUGE-1 = 0.351648, ROUGE-2 = 0.171429, and ROUGE-L = 0.274510. The data partitioning made LexRank more effective in selecting important sentences, thereby improving the quality of the summaries compared with summarizing the entire dataset at once. The ROUGE evaluation indicated that the generated summaries were reasonably accurate and informative.