Designing a Teacher Recommender System: A Thematic Literature Review of Teacher Evaluation Systems

Teacher evaluation systems are limited in their ability to provide numerical ratings, often failing to analyze qualitative feedback to provide teachers with valuable insights to enhance performance. This paper conducts a thematic literature review of teacher evaluation systems and tools in articles in Google Scholar, IEEE, and Proquest databases between 2014 and 2024 to determine the most appropriate sentiment analysis (SA) and topic modeling (TM) algorithms for analyzing student feedback. The review of 48 articles found that a lexicon-based SA approach, specifically VADER with a customized Filipino lexicon, offers a robust and practical solution for sentiment detection in a multilingual context. For TM, Latent Dirichlet Allocation (LDA) with human intervention is the recommended approach, providing a balance between thematic granularity and computational feasibility. The efficacy and efficiency of both algorithms are found to improve by increasing the size of a domain-specific corpus of words. Based on these findings, the paper proposes the design of TeachAIRs. This teacher recommender system includes word cloud visualizations, sentiment scores per topic, and, most critically, actionable insights derived from the integrated analysis. The development of this system is highly recommended to provide teachers with valuable and constructive real-time feedback, ultimately enhancing teaching practices and improving student learning outcomes.