Infret
TL;DR
Infret is an educational information retrieval tool that I developed for my master's research project. Through evaluation in multiple university classes, Infret has proven to be a highly effective visualization and exploration tool that is still in use today. My analysis of the usage data and feedback surveys has enabled me to better understand how Infret helps students learn information retrieval concepts and meet their educational needs. Infret was later extended by a master's student under my mentorship with coding features which enable users to write custom IR metrics. It is also in use as part of an Information retrieval class for multiple years and helps students understand various information retrieval concepts.
3. Nov 2023
The Aim
The idea behind Infret was to provide an educational information retrieval application for computer science students. The initial aim was to cover basic concepts of information retrieval such as term weighting and gain a better understanding of how such solutions help students grasp the information retrieval topic. To do so I first developed the application using Hapi, MySQL and Angular. Next, I conducted a user study as part of which I explored how the tool is perceived by its users through usability, experienced emotions and some direct feedback as well as how it helps the users learn concepts better. The analysis of results was done using traditional analysis tools such as Jupyter, Pandas, NetworkX and Plotly. In addition to analysing the survey data I analysed interaction data collected during the usage of the tool. Some of the findings and lessons learned are discussed in the next section.
Findings and Lessons Learned
Research perspective: Throughout multiple studies this project showed how introducing more interactive materials into a class helps students better understand concepts as well as motivates students to explore various new topics. Measuring usability in addition to asking for explicit feedback about Infret also proved to be a very simple and effective way of identifying the severity of UX problems.
Technical perspective: While Infret was built with modularity in mind and the code was split into logical components, it was not containerized . Furthermore, the choice of a Node-based framework for the backend was a sound decision since it meant that people working on the project only needed to learn a single language. Despite Javascript/TypeScript not being the primary language of choice for data applications in the context of educational apps it provides an environment that is easily extensible as well as a variety of libraries that can be used for NLP, IR and related areas. Dealing with interaction data also proved to be interesting and challenging since the initial model of the tracked interaction data also influenced the later analysis of the data. To analyse this data I split it into three granularity levels: areas, sub-areas and actions. Depending on the questions one wants to answer they should choose one of these three granularity levels. In my case, these were the area and sub-area levels which enabled me to gain an insight into how students navigated between multiple concepts presented in Infret.
Papers
In addition to the creation of Infret, this project also resulted in the two papers listed below as well as a master thesis as part of which I acted as a co-supervisor.