The ability to engage in a process of argumentation is essential for human beings. Humans use argumentation to communicate and defend their justifiable positions (or opinions), to understand new problems and to perform scientific reasoning. The process of argumentation has been studied for centuries and in different research domains, namely: philosophy, psychology, linguistics, computer science and artificial intelligence.
The aim of argumentation mining is the automatic detection and identification of the argumentative structure contained within a piece of natural language text. In a general form, arguments are justifiable positions where pieces of evidence (premises) are offered in support of a conclusion. The ambiguity of natural language text, different writing styles, implicit context and the complexity of building argument structures are some of the challenges which make argumentation mining very challenging.
By automatically extracting arguments from text, we are able to tell not just what views are being expressed, but also what are the reasons to believe those particular views. Therefore, argumentation mining has the potential to improve some research topics such as opinion mining, recommender systems and multi-agent systems.
In the ArgMine project, we aim to combine techniques from computational linguistics and machine learning with argument structure and rhetoric theories in order to automatically extract the argumentative reasoning included in textual documents, such as opinion articles, written in the Portuguese language. To address this challenging task, we envision and are developing the ArgMine Framework, which aims to integrate the process of creating a corpus annotated with arguments and the semi-automated process of selection and experimentation of different models and relevant features in different steps of the argumentation mining process.
Supervised machine learning techniques require a considerable amount of annotated data (in this case, arguments annotated in documents) in order to learn how to automatically perform this task. In order to make the annotation task easier and more intuitive to perform, we developed an annotation platform, with which we are creating the first argumentation mining corpus in Portuguese. The platform includes a drag-and-drop interface to annotate arguments from text and some starting material explaining how to build argument diagrams from text and introducing several argument diagramming techniques.
The main lines of research that are currently being explored in the ArgMine include: