Utilizing Latent Semantic Analysis to Provide Automated Educational Support

The traditional in-class methodology was developed for small classrooms of 15-20 students. Low student to teacher ratios, typically under 20 students per teacher, have been preferred and recommended to maximize student achievement, engagement, and retention from research starting in the 1970's [1] [2] [3]. Actual classroom sizes for K-12 vary depending on a variety of factors [4]. Today, some undergraduate Engineering courses consist of more than ten times that many students: some who are interested, some who just want a passing grade, and others who are not yet ready for college and do not properly prepare to study material. In fact, according to a national survey consisting of 560 colleges and universities in 2016, 20% of first-year college students had difficulty learning and getting help with coursework [5] [6]. As classroom sizes increase and varying levels of experiences of students, this situation will only exacerbate existing problems and deficiencies utilizing current teaching methodologies and tools. An automated tool that can provide similar recommendations would free up all that time and allow for more meaningful discussions. Also, students would save hours individually in terms of getting stuck, waiting for responses, and then spending time to get back to where they were later when they got stuck. This is potentially even more beneficial for students who do not typically ask questions when they get stuck, hoping that attending lecture or discussion will answer their questions. Utilizing latent semantic analysis (LSA), a natural language processing algorithm, recommendations can be created through mathematical searching and categorizing sources using singular value decomposition (SVD). The automated tool can pre-emptively suggest additional reading and viewing material, allowing the student to continue their studies without a long wait interval.