Modelling text semantics
Information science and technologies. Bulletin of the ACM Slovakia chapter, (2017)bibtex: vsajgalikmodelling.Abstract
In the dissertation, we focus on modelling text seman- tics. We identify two sub-goals, which aims at modelling abstract text semantics. While the first sub-goal is ori- ented on modelling the general text semantics, the second sub-goal is focused on the discriminative semantics, which can be of more information value. Besides proposing new methods to fulfil these sub-goals, we also examine a prac- tical application of our proposed method of discriminative keyword extraction.
Our contribution can be split into three parts. First, we propose a method to model abstract text semantics via key-concepts and show how it improves over standard key- word extraction methods. As a second contribution, we propose a method to model discriminate abstract text se- mantics, which is based on categorised text documents. We show how better representation of text semantics can improve over state-of-the-art methods in text categorisa- tion even with traditional keywords. Finally, we propose an approach to modelling user interests using our method of discriminative keyword extraction, which is evaluated on real-world noisy data in diverse domains.