CLASP
The Centre for Linguistic Theory and Studies in Probability

HotAM: an Argument Mining Framework

Argument Mining is a complex task, and to solve it, one has to create solutions that deal with several different subtasks. Existing research is comprised of a few end-to-end attempts as well as combinations of solutions to one or more of the subtasks. To determine which is the best end-to-end solution, one must combine models solving individual or combinations of subtasks with one another. As I set out, I found myself with code in different languages and different code bases. I saw an opportunity to unify the existing research into a single Python framework not only to create a better setting for experiments but also to make research in Argument Mining easier and more accessible.

This tool includes preprocessing, feature extraction, deep learning training setup, SOTA models, visualisations, experiment and model logging, and other features.

In this seminar, I will introduce you to Argument Mining, its challenges and my framework for Argument Mining; HotAm.