Verification of Disinformation through Information Retrieval and Natural Language Inference

The rise of machine learning and natural language processing (NLP) has opened a range of possibilities for automating and optimising tasks across many domains. Disinformation, a significant social issue, presents a challenge that can be addressed using these new tools. In this thesis, we propose an efficient and flexible architecture to automate and assist in the task of information verification, that is, validating natural language claims against a database of established ground truths. We have successfully trained our system on the FEVER dataset and obtained promising results, both in evidence retrieval and entailment, proving the viability of our approach.