Andreea Iana Defended her PhD
In her thesis, which was in part an outcome of the ReNewRS project, Andreea has made major contributions in three areas:
- A careful dissection of existing Neural News Recommender Systems, analyzing their common building blocks and developing the modular NewsRecLib framework which allows for assembling them in a plug-and-play fashion. Moreover, she has shown that common design decisions in Neural News Recommenders can be replaced by simpler alternatives at equal or even better performance, with, at the same time, smaller memory footprint.
- The support for diverse recommendations, where diversity can come in multiple facets. This lead to the development of the MANNeR framework, which allows for training diversification modules independently, and combining them on the fly for a diversified news recommendaiton.
- The extension of the common mono-lingual paradigm to support multi-lingual news recommendation, including the provision of the novel xMIND dataset for evaluation, and an extensive set of experiments also simulating different multi-lingual news consumption patterns.
The evaluation committee, consisting of Julia Neidhardt (TU Vienna), Goran Glavaš (University of Würzburg), Heiko Paulheim (University of Mannheim), and Simone Ponzetto (University of Mannheim), unanimously recommended to pass her dissertation with summa cum laude. Congratulations, Andreea!