Prediction of schizophrenia with
- NGram Tracing (https://www.researchgate.net/publication/329795318_Attributing_the_Bixby_Letter_using_n-gram_tracing) and
- Word Graph Analysis (https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0034928) which were compared to a baseline of Bag of Words with Naive Bayes and Support Vector Machines. Both methods (NGram Tracing and WGA) tend to overfit on the small amount of data). The full report can be found in the file Schizophrenia_prediction.pdf of this repository.
This was a graded project for the Individual Module (IM, 15 ECTS) of the Cognitive Systems Master's program at the University of Potsdam (Germany), supervised by Professor Walter Daelemans at the research group CLIPS at the University of Antwerpen (Belgium) and Professor Manfred Stede from the University of Potsdam (Germany).
The data was collected by CAPRI at the hospital of Antwerpen (Belgium) and is under non-disclosure.
The project was presented at
- the CLIN30 conference in Utrecht on January 30st, 2020: https://clin30.sites.uu.nl (poster presentation)
- the Poster Slam of the University of Potsdam on April 29th, 2020, held online via Zoom