There have been a large number of approaches to automated argument mining. The mainstream approach consists in applying machine learning (ML) techniques on manually annotated corpora to obtain a tool which automatically identifies argument structure in texts. In these cases, a corpus and its annotations determine the capabilities of the resulting tool. As an alternative to ML, some tools have been developed that rely on hand-crafted rules. For example, J. Kang and P. Saint-Dizier [KAN 14] present a discourse grammar implemented as a set of rules and constraints.
In this chapter, we present some systems that have been developed to automatically recognize the argumentative structure and the arguments in texts and to classify the detected arguments according to their type (e.g. counterarguments, rebuttals).
7.1. Application domains for argument mining
Argument mining serves the purpose of identifying why a speaker/writer holds such opinion. Some argument mining systems aim at describing argumentation (i.e. simply discovering arguments and, possibly, the relationships between several arguments), while another application of argument mining is to evaluate arguments (i.e. determining whether they are sound or to assess their influence in decisions). The most obvious application for argument mining systems is the analysis of the argumentative structure of a text (whether it is written or transcribed from oral data) to find the conclusion ...