Abstract In some sense, the question about the quality of an argument is the ‘ultimate’ one for argument mining. Being able to reconstruct argumentative structures from a text (Chaps. 4 – 5 ) and learning about the perspectives that arguments convey toward a topic (Chap. 6 ) lays the ground to make arguments processable by machines. However, nearly all practical applications of NLP research on argumentation require some distinction of good from bad arguments, in order to focus on the strongest arguments or to identify weak arguments. Argument search engines aim to rank the best arguments highest, debating technologies aim to choose arguments that help convince their opponent or audience, argumentative writing assistance tools aim to spot the weak points to give feedback on, and so forth. In this chapter, we therefore discuss how to assess argument quality with computational means.
Abstract Chapters 4 and 5 have presented the core of argument mining, that is, the identification of argumentative structure in natural language texts. While this information constitutes the basis of any application dealing with arguments, many applications also need knowledge about the perspectives that mined arguments convey toward the controversial issues discussed. For example, existing argument search engines separately display pro and con arguments on an issue. Debating technologies should frame arguments in a way that supports their side convincingly, and the effectiveness of arguments may depend on whether the values they encode match those of their audience. In this chapter, we review NLP methods that assess the perspectives of arguments or argumentative texts.
Abstract This chapter discusses tools for working with arguments as well as applications that use or could use argument mining. In our usage, tools process arguments as input or output for well-defined tasks within argumentation theory; they may be relevant for tasks such as creating corpora or assessing argument quality. By applications we mean complete argumentation systems designed to support or interact with human users to achieve a specific goal that substantively involves argument or controversy in some way.
Abstract Reasoning is the heart of argumentation—so much that some researchers might equate the terms reasoning and argumentation . This is because, conceptually, an argument manifests a reasoning process (or process step) in natural language: the relation that it defines from one or more premises to a conclusion instantiates the act of giving a reason, or reasons, to support a claim. In a broader sense, though, argumentation comprises more than just reasoning. As the previous chapters on perspective and quality assessment have emphasized, subjective viewpoints and perceptions are inherent to argumentation, and rhetoric and dialectic are important in any real-world argumentation setting. Thus, reasoning can be understood as the part of argumentation that describes how people infer new (or hidden) information from given information through the use or exchange of arguments. As such, reasoning is where the NLP tasks of analysis and synthesis meet. In this chapter, we discuss some main strands of NLP research that model key aspects of reasoning, given previously mined or otherwise structured arguments.
