Argumentation mining in short text: detecting argumentative information in real-life settings

The recent technological leaps of artificial intelligence (AI) and the rise of machine learning (ML) have triggered significant progress in a plethora of natural language processing (NLP) tasks. One of these tasks is argumentation mining which has received significant interest in recent years and is regarded as a key domain for future decision-making systems, information retrieval mechanisms, and natural language understanding problems. In the meantime, the developments beyond Web 2.0 have transformed the means of communication and information exchange, promoting shorter bursts of text without solid argumentation. Modern research questions and challenges indicate a need to develop innovative research methods and mechanisms that enhance the trust and the explainability of the AI services enabling the transferability of knowledge between tasks and domains not only on carefully constructed datasets but also in real-life settings. The main objective of this thesis is to provide a better understanding of natural language in realistic scenarios enhancing the trust in the NLP systems through the task of argumentation detection. Detecting argumentative segments in short text is a crucial step towards a deeper understanding of human language because it delves deeper into the reasoning process and quantifies previously unexplored qualitative aspects. The integration of qualitative aspects through prior knowledge into NLP pipelines has the potential to offer a new perspective and increase the trust in the outcome of AI solutions. This thesis reviews the task of argumentation detection, defines the theoretical foundations for agile argumentation frameworks, offers an annotated dataset to the research community, presents the benefits of integrating symbolic AI into hybrid solutions, and examines the suitability of contextual embedding for the task of argumentation detection.