Argument Mining

6Annotation Frameworks and Principles of Argument Analysis

Research in linguistics mostly relies on empirical analyses to characterize or understand a discourse feature. Corpus analyses are particularly useful because they provide detailed and thorough information about peculiar discourse characteristics.

In computational linguistics, manual annotations of naturally occurring examples have been used as a source of training material. This task consists in annotators manually assigning an interpretation to texts.

Argumentation theorists carry out annotations of texts to detect, identify and evaluate arguments. This analytical task is necessary to understand and characterize facets of human reasoning. Manually annotating argumentative texts, indeed, allows the detection of argumentative components which may not be anticipated. It is also necessary to manually perform annotations prior to building systems that can automatically replicate them afterwards. Moreover, argument analyses can be used to provide examples and counterexamples of argument features in order to feed a program that will automatically detect them or to develop rules and grammars. In other words, annotated data are used as training sets to test and develop argument mining systems.

In this chapter, we describe some annotation frameworks (or models) that have been used to analyze arguments, with an emphasis on the different argument elements that can (or must) be identified and analyzed. We also emphasize the need ...