Doctoral Candidate at UNICAEN


Søren Fomsgaard


Coming from computational linguistics and philosophy technology, I am interested in how linguistic patterns  or styles differ between humans and machines.

Textual online content generated and/or propagated by non-human agents has become an increasingly present phenomenon on social media platforms. When deployed with malicious intent, the behavior of this kind of agent can have impacts that are either directly harmful to people that are targeted by this behavior, for example, through toxic language use, or impacts that are more indirectly harmful, for example, by intervening in political discourse.

With the advent of contemporary large language models, the near future may feature more bots that employ increasingly sophisticated human-like harmful language online. Therefore, being able to detect linguistic patterns and idiosyncrasies will play a role in identifying harmful textual content that is not only spread but also generated by machines.

My research interests include authorship analysis, computational sociolinguistics, and the philosophical issue of what defines and differentiates human, non-human, and hybrid forms of agency. I joined the HYBRIDS project because of its focus on interdisciplinary research, interaction, and collaboration, both in its individual research tracks and at the network level.

ESR6 Individual Project

Detection of toxic bots in Twitter by analysing textual content

Research Objectives:

AI-Generated Text Classification for Detecting Political Bots

To design and develop classification strategies to detect AI-generated texts applied to detect different types of political bots that are able to influence election campaigns.

Twitter Data Collection for Campaign Analysis

To crawl and scrape Twitter data for a specific campaign and for a specific period.

Linguistic Analysis of Bot-Generated Messages

To analyse the linguistic content and relevant features of the messages sent by bots by focusing on main topics, extreme opinions, language patterns, etc.

Taxonomy Development for Bot Classification

To make a deep taxonomy of bots by considering different aspects, such as their aims, attitudes, linguistic register, or topics.

Expected Results:

Bot Detection and Classification System Development

Development of a system for bot detection and classification.

Fine-Grained Annotated Election Campaign Dataset Creation

Creation of a benchmark dataset related to a specific election campaign, provided with fine-grained annotations of different types of bots.

Comprehensive Evaluation Leveraging Knowledge and Datasets

Evaluation of the classification system by making use of the knowledge (not only the previously created dataset) acquired from the analysis.

Application to Diverse Use Cases: Brexit, Immigration, Climate, and Elections

Application on use cases related to European questions (Brexit and Euroscepticism), immigration, and climate emergency, as well as domestic political elections.


In our pursuit of academic excellence, HYBRIDS Doctoral Candidates are guided by a dedicated team of supervisors. Comprising the Main Supervisor, Co-Supervisor, and Inter-sectoral Supervisor, this team of professionals offers a wealth of knowledge, mentorship, and interdisciplinary insights.

Main Supervisor

Dr. Gaël Dias

University of Caen Normandy (UNICAEN)


Dr. Berta Garcia Orosa

University of Santiago de Compostela (USC)

Inter-sectoral Supervisors

Dr. José Ramon Pichel

Factoría Software e Multimedia (IMAXIN)

Mr. Ettore Di Cesare

Fondazione OPENPOLIS

Get in touch with us