Annotating videos that teach MS excel and predicting mouse / keyboard actions

Abstract

This paper presents a novel methodology for developing a structured taxonomy of Excel train-ing videos using natural language processing techniques and the PageRank algorithm. YouTube videos with closed caption transcripts were selected and annotated, extracting the main intentions of each video. The implementation leverages on TF-IDF vectorisation for key-word extraction, integrating it with the PageRank algorithm to rank these keywords based on their importance within the dataset. The resulting taxonomy of these keywords was created based on keywords that passed a ranking benchmark score. The taxonomy provides a founda-tion for the consistent categorisation of Excel educational resources. This allows further re-search to be built on structurally and steadily, circumventing the case of an overload of data with no avenues to search and categorise it. Future works could involve using more advanced techniques to extract relevant keywords, and expanding the taxonomy to be more diverse and to add depth to the categories.