This project's context is in the recent field of Argumentation Mining (AM), which deals with the automatic recognition and characterization of argumentative snippets. Relying on Argumentation Theory (AT) aims to offer an automated approach towards analyzing the structure and purpose of arguments.
This project aims to enhance AM by using ideas and approaches from Natural Language Processing (NLP). This will provide potential better results relying on the big annotated datasets available for general NLP tasks while adopting a transfer-learning to better adapt to the AM tasks. This combination is expected to foster models' attention to specific characteristics of the argumentations.
The final results will automatically measure the quality of existing implicit representations for argument mining and consider new implicit representation learning algorithms to capture the fine-granular properties of argumentation. The expected outcomes are an SW demonstrator and a scientific publication.