Scientia Potentia Est—On the Role of Knowledge in Computational Argumentation

Scientia Potentia Est—On the Role of Knowledge in Computational Argumentation Anne Lauscher1∗, Henning Wachsmuth2∗, Iryna Gurevych3, and Goran Glavaˇs4 1MilaNLP, Bocconi University, Italy 2Department of Computer Science, Paderborn University, Germany 3Ubiquitous Knowledge Processing Lab, TU Darmstadt, Germany 4CAIDAS, University of W¨urzburg, Germany anne.lauscher@unibocconi.it, henningw@upb.de, gurevych@ukp.informatik.tu-darmstadt.de, goran.glavas@uni-wuerzburg.de Abstract Despite extensive research efforts in recent years, computational argumentation (CA) re- mains one of the most challenging areas of natural language processing. The reason for this is the inherent complexity of the cogni- tive processes behind human argumentation, which integrate a plethora of different types of knowledge, ranging from topic-specific facts and common sense to rhetorical knowledge. The integration of knowledge from such a wide range in CA requires modeling capabil- ities far beyond many other natural language understanding tasks. Existing research on min- ing, assessing, reasoning over, and generating arguments largely acknowledges that much more knowledge is needed to accurately model argumentation computationally. However, a systematic overview of the types of knowledge introduced in existing CA models is missing, hindering targeted progress in the field. Adopt- ing the operational definition of knowledge as any task-relevant normative information not provided as input, the survey paper at hand fills this gap by (1) proposing a taxonomy of types of knowledge required in CA tasks, (2) systematizing the large body of CA work according to the reliance on and exploitation of these knowledge types for the four main research areas in CA, and (3) outlining and dis- cussing directions for future research efforts in CA. 1 Introduction The phenomenon of argumentation, a direct re- flection of human reasoning in natural language, has fascinated scholars across societies and cul- ∗Equal contribution. tures since ancient times (Aristotle, ca. 350 B.C.E./ translated 2007; Lloyd, 2007). The computational modeling of human argumentation, commonly re- ferred to as computational argumentation (CA), has evolved into one of the most prominent and at the same time most challenging areas in natural language processing (NLP) (Lippi and Torroni, 2015). CA encompasses several families of tasks and research directions, the main ones in NLP be- ing argument mining, assessment, reasoning, and generation. Although it bears some resemblance to other NLP tasks, such as opinion mining and natural language inference (NLI), it is widely ac- knowledged to be of much higher difficulty than the other tasks (Habernal et al., 2014). While opin- ion mining (Liu, 2012) assesses stances towards entities or controversies by asking what the opin- ions are, CA provides answers to a more difficult question: Why is the stance of an opinion holder the way it is? In a similar vein, while NLI focuses on detecting simple entailments between statement pairs (Bowman et al., 2015; Dagan et al., 2013), CA addresses more complex reasoning scenarios that involve multiple entailment steps, often over implicit premises (Boltuˇzi´c and ˇSnajder, 2016). CA targets reasoning processes that are only partially explicated in text. Its mastery thus re- quires advanced natural language understanding capabilities and a substantial amount of back- ground knowledge (Moens, 2018; Paul et al., 2020). …

Content truncated to 500 words