Trustful Natural Language Processing

Kristian Miok

Recent advances in Natural Language Processing (NLP) have achieved great success. This is especially the case for large pre-trained language models that nowadays provide state-of-the-art results for most of the NLP tasks. However, one of the main challenges that pre-trained language models still have is providing completely wrong results with high confidence. To address this challenge, NLP researchers work on various frameworks that could overcome this problem developing methods that could be named Trusful NLP methods. In this presentation, we will talk about the background of this challenge and indicate possible solutions and further directions for Trusful NLP modelling as a field.