Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

The thesis presented here investigates how Donald Trump employed conspiracy theories to manipulate public opinion. It is concerned with the representation of China and COVID-19 in his tweets and speeches, as well as the conspiratorial notion of a ‘deep state’. By way of data analysis and data visualisation, it spotlights spikes in conspiratorial discourse with regard to the aforementioned concepts. This analysed and visualised data is then further analysed in a more qualitative way.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This report is the result of an internship under the BESOCIAL research project, hosted by KBR. The main objective was to develop a case study exploring the use of contextual data extracted from the Twitter API and what insights could be extracted from applying this methodology to the members of four of the Belgian legislative chambers: The regional parliaments of Flanders, Wallonia, and Brussels, as well as the Belgian chamber of representatives. The present report documents all the steps taken during this exploratory exercise as well as some conclusions highlighting the split in how representatives from different Belgian regions, institutions and political parties use Twitter. The main outcome may be the confirmation of the deeper adoption of the social network by the representatives of the Flemish region, and how different political parties present different profiles in what respects to several dimensions of the Twitter experience. Aside from that, we can also notice some divergences on the usage patterns across parties and institutions. However, the result of this report is not an explanation of the reasons behind such phenomena but the mapping and identification of vectors from which Twitter contextual and meta data can be used for the analysis of cleavages between user groups in Twitter.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Computational techniques play an increasingly important role in analyzing large volumes of textual data. This research explores the methods to analyze the tweet dataset of Death of Queen Elizabeth, mining the social networks of the audience engaged in the communication and their opinions. This internship program is hosted by the Faculty of Art on KU Leuven, employing the dataset provided on iCANDID, a datahub platform by KU Leuven(LIBIS). The dataset is retrieved with tweets containing one of the three hashtags: #queenelizabeth #abolishmonarchy #queueforthequeen. While observing the dataset, the methodology is designed from two aspects of research: Social Network Analysis (SNA) and Text Mining. The SNA technique applied Python and Gephi to prepare data (nodes and edges) and visualize, respectively. In comparison, the approach to conduct text analysis was first to preprocess text data with common natural language processing (NLP) pipeline, then generate word clouds, and extract topics by the Latent Dirichlet allocation (LDA) model, all of which was mainly performed in Python. Overall, the results offer an overview of public opinions related to the Queen's death and significant users in opinion communication as well as their views. Regarding SNA, there are three social networks found from this dataset, which are co-occurring hashtag network, mention network, and reply network. Furthermore, five influential users in mention and reply networks are selected to research their tweets' content. Four of them, @Arad87709987, @nika6547, @talhamuneeb7700, and @Brijend84228427, are highly suspected to be chatbots, while the remaining one @HighwaySqueak, who has the strongest influence on information transmission in the network, claims strong opposition against the monarchy. In terms of text analysis, public opinions are primarily about condolence for the Queen's passing, the Queen's funeral, Blessings toward the Queen, discussion about King Charles, political views of abolishing the monarchy, and cryptocurrency. Surprisingly, the last topic is unrelated to the general topic of Queen Elizabeth.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

The present work aims to discuss the use of CLIP (Contrastive Language-Image Pre-Training) for the exploration of a until now unexploited digitised collection of historical postcards. In the Introduction, we present recent work on the computational exploitation of Humanities collections with computer-vision models. Secondly, we describe the functioning of CLIP and its recent application to case-studies from the Humanities. Third, we introduce the main features of the postcards collection. In the first Chapter, we test whether CLIP can be successfully used to retrieved churches and other classes of buildings. We assess whether the metadata of the collection can be used as gold data for evaluation and whether the exploitation of the data using a vision-only approach and of the metadata using a text-only approach outperforms the use of CLIP. We conclude that, in case we don’t apply any finetuning, CLIP functions better than the models used for comparison. In the second chapter, we focus more precisely on the comparison between the vision-transformer ViT and CLIP for the scope of classification using ImageNET labels. We conclude that, while still showing good performances, CLIP appears slightly unreliable when used to classify images using a broad array of labels. In the last Chapter, we investigate the potential of CLIP for extracting abstract or more generic concepts from the images. In particular, we explore the potential match between the interest of researchers working on postcards collections and the queries that are made possible by CLIP and not by traditional computer-vision models and classifiers.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This study thoroughly explores the affordances and challenges of electronic literature on Twitter through both theoretical and empirical analysis. The first part of this thesis establishes that electronic literature roots its origins in older literary practices of the avant-garde movements. However, it also highlights that the history of electronic literature is inextricably tied to the history of computing, networking, and their social adoption. The emergence of new digital platforms on which to conduct literary experiments enabled writers to extend their creativity beyond the print medium, opening the artistic sector to a whole new network. Examining the case of Twitter as one of these new creative platforms, this first part delves into the characteristics of the network. Despite its initial design without any literary focus, Twitter has evolved into a significant platform for electronic literature, attracting both professional and amateur writers. Features such as hashtags, bots, community engagement, and widespread distribution have fostered aesthetic and conceptual innovations in electronic literature. Literary bots, at the heart of the empirical analysis conducted in a second part, are as a compelling manifestation of electronic literature on the platform. The second part is based on the first phase of a project aiming at identifying and classifying electronic literature: project LabEL. Detailing the progress of the work, this parts empirically looks at the difficulties encountered through the iterative process of shaping classification labels for the procedures used by a set of English literary Twitter bots. In the discussion of the theoretical findings and empirical considerations, this thesis draws two main conclusions. Firstly, the study of literary Twitter bots reveals a strong influence of avant-garde methods and techniques, both in their automated procedure and in their qualitative output content. The procedures employed by these bots, such as transformation or templates, and the incorporation of randomness as a creative process, reflect the experimentation and reimagining of literary practices of movements like Dada, Surrealism and the Oulipo. The digital media therefore provides new ways of exploring older procedures. Secondly, the research points out the major difficulties encountered in studying literature on Twitter. The dynamic nature of the platform, the absence of static genres, and the limitations imposed by Twitter's APIs pose significant obstacles. The suppression of accounts, notably as part of anti-bot campaigns further complicates the data collection and the construction of comprehensive databases for analysis. Nevertheless, these challenges underscore the urgent need for projects like LabEL, which not only investigate the remediating of older procedures in new media but also contribute to the preservation of born-digital heritage.