Between 2019 and 2023 I did my PhD at the Knowledge Media Institute (Open University) under the supervision of Harith Alani and Alistair Willis.

In my work I propose a methodology to automatically compare persuasion means in news articles across different political orientations. This lies at the intersection between multiple research areas: corroboration and omission analysis, persuasion techniques extraction (propaganda and sentiment), political leaning and topic analysis.

In my work I make use of several NLP techniques and tools, such as document embeddings, clustering algorithms, transformer models for sequence tagging, supervised learning through neural networks.

The following materials are publicly available:

Step By Step

The direction for the thesis was refined during the years.

In my first year I focused on analysing of how different articles present the same events, by using different emphasis and selecting different details, aiming to produce a comparative analysis. You can see more in the video below or take a look at the poster sent to the OU poster competition 2020. A more detailed explanation of the initial objectives can be seen in the position paper that has been presented at the Text2Story workshop.

This direction can be seen in the Upgrade Report

Then in my second year I shifted my goals more towards the automatic detection of propaganda techniques, and how different news sources use them in specific ways.

And in my third year I analysed how these techniques vary across topics and how the current approaches for propaganda detection are specifically targeting extreme-right propaganda.

In my fourth year, my main activity was to refine the analyses and to report everything in the PhD Thesis.