Supervised Content Visualization of Scientific Publications: A Case Study on the ArXiv Dataset

摘要

A supervised approach to visualization of collections of scientific documents is presented. We have implemented a text classification module, which leads to class probability estimations, along with a dimensionality reduction technique which represents each class in the 2-D space. Integrating those two procedures, any collection of unlabelled documents can be visualized. The arXiv dataset has been adopted for training the classification and visualization modules. We demonstrate the system's functionality on a corpus of automatically detected publications of particular EU FP7 funding categories.