Argumentation mining (or argument mining) is the research area aiming at extracting human arguments and their relations (primarily) from text, with the ultimate goal to understand humans’ arguments in argumentative texts. Nonetheless, the progresses thus far often ignore the computational models of structured argumentation formalisms which are well studied in the knowledge representation and reasoning (KRR) community. This means that there is a still big research gap if ones would like to automatically construct argumentation-based knowledgebase systems from text. In this paper, we investigate this problem and formally define important tasks for argument mining from the views of machine learning and KRR. Due to current progresses of KRR, there could be a plethora of tasks depending on targeted structured argumentation frameworks. This paper focuses on a well-known assumption-based argumentation (ABA) by defining relevant tasks for mining ABA from text and then accordingly build an annotated corpus in hotel reviews from Booking.com. Finally, we evaluate the in-context learning capabilities of large language models (i.e. GPT-4o) for an ABA’s construction based on our annotated hotel reviews’ corpus, demonstrating the proposed framework is promising in practice.