As a media scholar primarily focused on sociology, the opportunity to participate in the SoBigData++ TNA program at the Department of Computer Science, University of Sheffield, presented both a formidable challenge and a unique avenue for improving analysis skills. The study of media, and particularly online disinformation, has reached a point where remaining confined within the boundaries of one's own discipline is no longer feasible. The necessity for cross-disciplinary skills has become paramount. From this perspective, my visiting experience at So Big Data ++ was immensely rewarding.

Upon arrival, I was immediately made to feel welcome by my host and their staff. This environment fostered a genuine opportunity to learn tools and methodologies vastly different from those I was accustomed to. The work began with clarifying some theoretical aspects, with the primary goal of the project being the detection and verification of problematic information on social media.

A crucial step in our work involved reaching a consensus on what constitutes a "narrative," the foundational element of detection (what exactly needs to be detected?). We quickly realized that the concept of narrative varies significantly across disciplines, swinging ambiguously between the notions of topic and framing. Establishing a shared definition of narrative—a task that necessitated several seminars—was our first milestone.

With a cohesive understanding of narratives, we embarked on developing techniques for their detection.

The work was primarily focused on three key activities. These activities were aimed at overcoming the barriers posed by linguistic diversity, the multifaceted nature of content, and the subtleties within narratives that make detection and verification a complex task. 

1) Automated Translation Scripts: A significant part of my work involved creating methodologies that allow to automatically translate texts from various languages into English. Given that disinformation often travels across different countries and languages, it's crucial to aggregate similar narratives even when they are in different languages. These instruments were crucial in breaking down linguistic barriers, enabling us to analyse and compare narratives on a global scale.

2) Multimodal Detection: The second area of focus was on developing capabilities for multimodal detection. This involved integrating different types of content, such as texts and images, to form a more comprehensive view of the information landscape. Since problematic information often utilises a combination of media to convey and strengthen narratives, being able to detect and analyse these multimodal elements was crucial for a more accurate identification of disinformation.

3) Pattern Identification in Narratives: Lastly, a key part of our approach was to identify common patterns within narratives, such as actors involved, causal attributions, and moral evaluations. Recognizing these elements helps in clustering similar narratives, facilitating the process of debunking and fact-checking.

These activities not only enhanced our ability to tackle disinformation but also underscored the importance of interdisciplinary approaches in this field. The combination of linguistic analysis, multimodal content detection, and narrative pattern recognition represents a holistic strategy for understanding and combating the spread of problematic information. 

Through this experience, I gained invaluable insights into the complexity of online disinformation and the innovative methodologies required to address it effectively. The collaborative environment and the diverse skill sets at the SoBigData++ TNA program were instrumental in advancing our collective efforts to safeguard information integrity on social media platforms.