Home > Published Issues > 2024 > Volume 15, No. 6, 2024 >
JAIT 2024 Vol.15(6): 693-703
doi: 10.12720/jait.15.6.693-703

Contextual Understanding of Academic-Related Responses Based on Enhanced Word Embeddings, Clustering, and Community Detection

Mary Joy P. Canon 1,*, Lany L. Maceda 1, and Nancy M. Flores 2
1. Computer Science and Information Technology Department, Bicol University, Legazpi City, Philippines
2. College of Information Technology and Computer Science, University of the Cordilleras, Baguio City, Philippines
Email: mjpcanon@bicol-u.edu.ph (M.J.P.C.); llmaceda@bicol-u.edu.ph (L.L.M.); nancy@uc-bcf.edu.ph (N.M.F.)
*Corresponding author

Manuscript received January 11, 2024; revised February 14, 2024; accepted February 26, 2024; published June 5, 2024.

Abstract—In strengthening educational policy, it is crucial to understand recipients’ feedback and experiences. This paper presents an innovative method for generating contextual analysis of qualitative responses from the beneficiaries of a free education program. We employed a two-tier clustering approach using the K-means algorithm and Louvain community detection, based on enhanced word embeddings and a bi-gram network. Our methodology extends beyond traditional text analysis by combining Word2vec and Glove embeddings, which captured the semantic meaning of words within the dataset. Through the application of the K-means algorithm, we identified five distinct clusters with a notable silhouette score of 0.3477, corresponding to themes of “Support and Educational Opportunity”, “Accessibility and Financial Relief”, “Gratitude and Satisfaction”, “Positive Evaluation with Suggestions for Improvement”, and “Program Effectiveness”. Further refinement of the clusters with the Louvain method, coupled with the use of a bi-gram text network instead of a uni-gram, achieved a higher modularity score of 0.637. While the transition from K-means clusters to Louvain communities resulted in slight thematic changes, it provided a more comprehensive view of the relationships among different response aspects. The two-tier clustering method highlights the methodology’s strengths and effectiveness in revealing hidden patterns and themes in the text responses. The findings not only highlight the strengths of the free education program in providing support to the beneficiaries but also reveal certain areas needing attention and improvement, which are crucial in policy development and enhancement.
Keywords—enhanced word embedding, clustering, community detection, text analysis, quality tertiary education, program evaluation

Cite: Mary Joy P. Canon, Lany L. Maceda, and Nancy M. Flores, "Contextual Understanding of Academic-Related Responses Based on Enhanced Word Embeddings, Clustering, and Community Detection," Journal of Advances in Information Technology, Vol. 15, No. 6, pp. 693-703, 2024.

Copyright © 2024 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.