Published Date

20

November 2025
Thursday

AIUB Students Achieve Best Paper Runner-Up Award at AII 2025 in Washington, D.C.

Events Date:
September 28
Year: 2025
Organized By:
Applied Intelligence And Informatics Lab
Venue:
Washington University of Science and Technology (WUST), Washington, D.C., USA.

The American International University-Bangladesh (AIUB) is proud to announce that a team of talented undergraduate researchers from the Department of Computer Science has secured the Best Paper Runner-Up Award at the 5th International Conference on Applied Intelligence and Informatics (AII 2025). The conference was hosted by the Washington University of Science and Technology (WUST) in Washington, D.C., USA.

The award-winning paper, titled “Overfitting-Aware Comparative Study of Sentiment Classifiers on Chat-Based Datasets,” was authored by Shovan Roy, Mahmudul Haque Shakir, Sunipun Seemanta, and Riya Das, all undergraduate students from the Department of Computer Science, AIUB. The research was supervised by Dr. Md. Saef Ullah Miah, Associate Professor, Department of Computer Science, and Dr. M. Mostafizur Rahman, Professor and Head, Department of Mathematics, AIUB.

This recognition is especially noteworthy as the paper was selected for the award from a competitive pool of 250 international submissions, highlighting the global impact and academic rigor of AIUB’s student research initiatives.

About the Research

The awarded study addresses a critical challenge in Natural Language Processing (NLP): chat sentiment classification, an increasingly important task as digital communication continues to expand. Chat data typically exhibit informal language, class imbalance, and noise, factors that often lead machine learning (ML) models to overfit, thus compromising generalization and real-world performance.

Through extensive experimentation on benchmark chat datasets, the study examined the interplay between model complexity, generalization ability, and classification accuracy. A key finding of the research is that the Naïve Bayes classifier, when fine-tuned with alpha = 0.01, outperformed all other models, achieving an accuracy of 88.03% and a weighted F1-score of 0.88. The results underscore the importance of informed model selection and overfitting mitigation strategies for achieving robust sentiment classification in dynamic, real-world chat environments.

This achievement reflects AIUB’s dedication to fostering a culture of innovation, academic excellence, and impactful research among its students. The university congratulates the research team and their supervisors for representing AIUB on an international stage and showcasing the high-quality research emerging from the institution.

AIUB remains committed to supporting student-led research initiatives and expanding opportunities for global academic collaboration.

Post Title Post Title