Speaker: Simon Reisinger (193-02 Computer Graphics)

Whenever knowledge workers need to do research, they collect information from different origins in different formats. It is not easy to keep track of one’s sources and organise them efficiently, especially when there are many. The aim of this master thesis is to implement a program to simplify the process. The user can manually mark texts and articles in the browser and the program will automatically arrange them. The program uses document embeddings and dimension reduction to find and visualise them closer together or further apart depending on their semantic similarity. The applied embedding technique usually tries to unify all input strings to a fixed-length feature vector. We use dimensionality reduction on the resulting high-dimensional feature space so we can display all papers (represented by thumbnails) on a 2D screen. The user will have the opportunity to interact with the visual representation of the text data, in the form of a thumbnail by moving them on the provided area. This feedback will help the model improve the future results. 




10 + 10
Supervisor: Manuela Waldner