I'm completing this project as part of the Microsoft Student Accelerator program.

The intent of this project is to create a tool that can be utilised alongside the Arxiv website, which collects research papers. As a first-year student with an interest in research, I have often found myself wanting to further my studies by reading research papers, but without completing further studies in my areas of interest, many research papers are impenetrable. I wanted to create a tool that could make it easier to find papers linked to a specific field of an academic domain, utilising the information available about Monash courses online to tailor this project to undergraduates in a similar position to me. Future extensions of the project could include implementation adapted for other universities, or with more granular analysis of the accessibility of content in research papers.

This project is uniquely suited for an Artificial Intelligence solution. Performing a complex analysis of features and content within a paper and categorising it based on the results would be a time-consuming and nigh-impossible task without training a computational model that can learn to make its own predictions. In particular, a clustering model for machine-learning provides an approach to this problem which allows the program to perform feature identification rather than the builder of the program, which results in a more precise feature analysis for predictions than one that is human-created would typically be.

I didn't actually get anywhere near finishing this project, so there aren't really any installation instructions I can provide. Running the cells within preprocessing.ipynb has an extensive list of requirements, but the imports are all very obviously listed, so if you'd like to investigate my preprocessing process, definitely give it a look and install the relevant pip packages! I should note that several of the cells were used to modify (shrink and edit) the original data sets, which I haven't uploaded due to file size and ultimate irrelevance to the project. This means that you won't be able to run the cell code yourself, which is probably for the best as some of it would take >10 minutes every run on my computer (and the preprocessing code I initially wrote took over 10 hours to run and didn't finish successfully). My documentation should definitely explain the processes involved in the preprocessing if you're interested!