Monitoring Learner Interaction With Multimedia Content Using Natural Language Processing

E-learning has enhanced the way students think and learn over time. Different studies show that students appreciate facilities offered through the e-learning process. It has also proven to be a very useful tool by bringing education closer to the learner, hence, learner-centred. A point to note about e-learning is the ability to offer a wide range of information access, storage and education curriculum. Institutions of higher learning moderate and harmonize this information between instructors and students. These institutions also advocate for different styles and types of content presentation. This study focused on three types of content namely: audio, textual and video content. However, monitoring learner-content interaction in open environments where each student has his/her own device posed a unique challenge to the instructors and the institutions at large. The study also emphasized on investigating content items in asynchronous e-learning environment. It proposed the use of a monitoring tool to evaluate the level of content interaction as well as their dominance. The study developed an artefact that collected learner-content interaction data and reported to the respective instructor. Natural language processing was used for text extraction as well as pre-processing. It was then applied to scale through words and sentences that the learner had interacted with and provided a word count as a form of textual content interaction report to the corresponding course instructor.

Session-based interaction has been suggested in the study as a method of data collection in the level of content interaction. This was possible through the use of the percentage formula suggested in this study. The methodology, was design science and it revolved around development of an information system - based artefact. A population of USIU-Africa students was suggested with a sample size of 50 random students, each taking 4 different lessons to provide a dataset of 200. The frame was between freshmen and senior students. Using a 95% confidence level and 50% population variance with error margin of 5, the study was confident that the provided dataset would produce an accuracy of 132.