The increasing attention in the field of argumentation has motivated the scholars in many other disciplines to look for argumentative structures in their corresponding resources. Arguments may occur in various contexts from debate forums on the Web to discussions among the jury members in the Court. Searching for argumentative structures in the huge amount of data available led to the rise of a new study in the field of computational linguistics called Argumentation Mining.
Argumentation Mining is the automatic process of extracting arguments from various types of resources. An Argumentation Mining pipeline consists of several natural language processing algorithms.
This pipeline is composed of several certain stages in many studies first of which is to detect the components of an argument. Conclusions and premises are considered the two components of an argumentative text in most studies. Different machine learning methods for classification such as support vector machines, logistic regression and neural networks have been applied in order to detect components in argumentative text.
The second stage is to find the relation between each of the argument components and also the detected arguments (for instance if they are supporting or attacking each other). This stage is called argument structure prediction. Studies conducting researches in this stage have used various algorithms such as Textual entailment suites, parsing methods alongside classification methods.
Argumentation plays is a key factor in the legal domain. Therefore, recently many studies have focused on argumentation mining from legal texts. One of the crucial steps is acquiring corpora in this area to evaluate the efficiency of the implemented machine learning methods. One of the most distinguished corpora for argument mining in legal text is the ECHR corpus which gathers a set of documents from the European Court of Human Rights.
The University of Luxembourg and I3S research center at CNRS – Sophia Antipolis are engaged in a collaborative research in Argumentation mining in Political debates. This research is now in the initial phase of building an annotated dataset from the transcripts available at Commission on Presidential debates from 1960 to 2016 presidential debates in the United States.