Argument mining is still in an early stage of development. It raises much more complex challenges than information retrieval or question answering. Answering Why-questions in question answering could be the closest related challenge, if we consider that its main aim is to retrieve justifications to a claim. In contrast with information retrieval and question answering, mining attacks or supports to a claim cannot, in most cases, be based solely on linguistic cues, as shown in Chapter 4, since they are not very frequent or central. When linguistic cues exist, they can be ambiguous and may introduce, for example, causes or factual information, besides attacks or supports. Argument delimitation is also a real challenge.
Argumentation is a very complex phenomenon, and argument mining is just an approximation or a simplification of some aspects of argumentation analysis in order to extract prototypical arguments or parts of them with a reasonably good accuracy. Argument mining also depends on the goal that is pursued, for example, detecting fallacies, finding arguments, detecting argumentation strategies, finding claims, finding support or attack for a given stance, among others.
5.1. Some facets of argument mining
Due to these complexities, most of the projects aiming at automatically detecting arguments in text focus on argument annotations in various manners. Reliable, stable annotation from experts is required to build automatic systems. ...