Semantic Web 0 (0) 1 1 IOS Press A Review of Argumentation for the Social Semantic Web Editor(s): Harith Alani, The Open University, UK Solicited review(s): Fouad Zablith, American University of Beirut, Lebanon; Simon Buckingham Shum, The Open University, UK; Iyad Rahwan, Masdar Institute, Abu Dhabi Open review(s): Jodi Schneider a, Tudor Groza a;b, and tention of scholars from fields such as artificial in- Alexandre Passant a telligence [137], communication theory [5], business a Digital Enterprise Research Institute, National management [88] and e-government [107]. At the University of Ireland, Galway, same time, argumentation researchers began establish- fi[email protected] ing the foundations for a World Wide Argument Web b School of ITEE, The University of Queensland, (WWAW) as “a large-scale Web of interconnected ar- Australia, [email protected] guments posted by individuals to express their opin- ions in a structured manner” [138]. Arguments on the Web can be used in decision- making contexts. Decision-making often requires dis- cussion not just of agreement and disagreement, but Abstract. Argumentation represents the study of views and also the principles, reasons, and explanations driving opinions that humans express with the goal of reaching a the choices between particular options. Furthermore, conclusion through logical reasoning. Since the 1950’s, sev- arguments expressed online for one audience may be eral models have been proposed to capture the essence of in- formal argumentation in different settings. With the emer- of interest to other (sometimes far-flung) audiences. gence of the Web, and then the Semantic Web, this model- It can be difficult to re-find the crucial turning points ing shifted towards ontologies, while from the development of an argumentative discussion, even one in which we perspective, we witnessed an important increase in Web 2.0 have participated. Yet on the Web, we cannot subscribe human-centered collaborative deliberation tools. Through a to arguments or issues, nor are there tools that support review of more than 150 scholarly papers, this article pro- searching for arguments. Nor can we summarize the vides a comprehensive and comparative overview of ap- rationale behind a group’s decision, even when the dis- proaches to modeling argumentation for the Social Seman- cussion took place entirely in public venues such as tic Web. We start from theoretical foundational models and investigate how they have influenced Social Web tools. We mailing lists, blogs, IRC channels, and Web forums. also look into Semantic Web argumentation models. Finally By providing common languages and principles to we end with Social Web tools for argumentation, including model and query information on the Web (such as online applications combining Web 2.0 and Semantic Web RDF [90], RDFS [1], OWL [3], SPARQL [2], Linked technologies, following the path to a global World Wide Ar- Data principles [18], etc.), the Semantic Web [20] is gument Web. an appropriate means to represent arguments and ar- gumentation uniformly on the Web, and to enable, for Keywords: Argumentation, Semantic Web, Social Web, Se- mantic Web, Ontologies instance, browsing distributed argumentation patterns that appear in various places on the Web. Indeed, re- searchers have shown that the Semantic Web can be 1. Introduction used for visualization and comparison in decision ra- tionale [102]. In recent years, the problem of representing large- In this context, this paper discusses research in mod- scale argumentation on the Web has attracted the at- eling argumentation as it relates to the Social Seman- 1570-0844/0-1900/$27.50 c 0 – IOS Press and the authors. All rights reserved 2 Schneider et al. / A Review of Argumentation for the Social Semantic Web tic Web [10,70,27], focusing on foundational models mentation (Section 4) from a variety of fields, compare of argumentation, their applications in the Social Web, them (Section 5), and present applications of these the- and on ontologies (as in computer science [69]). oretical models (Section 6). Subsequently we present In particular, our purpose is to investigate ontologies (Section 7) and compare (Section 8) Semantic Web and tools which may be useful for argumentation on models of argumentation. Then we move on to review- the Social Semantic Web, a field where the aforemen- ing tools: in Section 9 we highlight thirteen notewor- tioned Semantic Web technologies support Social Web thy features of Social Web argumentation tools, based [123] applications, while at the same time Social Web on a comprehensive analysis of thirty-seven relevant paradigms are used to generate Semantic Web data col- tools (see the Appendix for full details). Finally we laboratively and at large scale. This convergence aims conclude the paper in Section 10. at providing new and improved ways to integrate and discover data, following the vision of Social Machines provided by Berners-Lee [19], both on the Web and 2. Background in the enterprise [127]. In the context of argumenta- tion, this could help to aggregate arguments from var- 2.1. Argumentation ious websites — for instance a discussion starting on Twitter and followed up on a mailing list, later frozen Argumentation theory is the study of agreement, on a wiki once consensus is reached — thus providing disagreement, and of the dialogues and writing through new means to follow argumentative discussions on the which we convince ourselves and others of our points Web. This would enable an argument-centric view of of view [65]. Informal argumentation occurs through- the Web. out conversations, online and offline, often in conjunc- Moreover, the Social Web does not yet have widely- tion with persuasion or with joint decision-making. used argumentative ontologies, though this problem has been noted [67], along with the need for federation Even logically sound decisions may involve choices infrastructures [125]. Thus, in order to identify how based on values and preference judgements: people different argumentation models and tools can be used may agree on the facts of a situation yet disagree on for the Social Semantic Web, this paper offers a review the preferred outcome or decision to be taken. That is of more than 150 research papers on the topic, from vitally different from disagreeing on the facts of a sit- 1945 to 2011, from which we compare: uation (in which case more information is called for). We are concerned with argumentative discussions, – 14 theoretical models of argumentation which we take to be online, mainly textual messages – 14 Semantic Web models for argumentation (i.e. and discussions, in which subjective perspectives or ontologies) differences of opinion are important and relevant. – 37 tools for representing argumentation on the Groups may use online conversations and social media Web. to coordinate and support decision-making; such argu- mentative discussions can be found in many online dis- As the focus is on human-centered argumentation cussion fora such as standardization bodies’ listservs, [87], with the goal of improving access and providing Wikipedia editors’ wiki pages, and open source com- overviews and visualizations, this article will briefly munities’ IRC channels, bug reports, and listservs. In- mention, but not analyze, the agent-based argumenta- dividuals may also draw on online conversations and tion domain1 social media in order to form personal opinions and Following the introduction, we provide brief overviews clarify their own preferences, based on sensemaking of argumentation (Section 2.1) and of the Social Web and analysis of others’ experiences; such argumenta- (Section 2.2), then discuss requirements for support- tive discussions can be found, for instance, on product ing argumentation on the Social Semantic Web (Sec- reviews websites, political blogs, in patient advocacy tion 3). We next present theoretical models of argu- and support group discussions, and in brief anecdotes shared via microblogs. 1 See for instance the Argumentation in Multi-Agent Systems There are a variety of common argument structures (ArgMAS) Workshop series, in its ninth year in 2012. In particular, in the social context, Heras has presented argumentation work from [134]. A single premise may directly support a conclu- a social perspective using case-based reasoning and ontologies e.g. sion (as in (i) of Figure 1 on the facing page), but more the ArgCBROntology 2. [76,73]. commonly, they are combined to come to a conclusion. Schneider et al. / A Review of Argumentation for the Social Semantic Web 3 Yet the internal structure of these discussions – such as whether the participants agree or disagree, are con- tributing diverse ideas, or debating in circles – is still not represented in SIOC. Capturing such underlying arguments would be valuable, and research is begin- ning to address this for instance by identifying argu- ment schemes used in Amazon reviews [75,208] and by modeling the speech acts in Twitter conversations [148]. Yet infrastructure for argumentation on the So- Fig. 1. Common argument patterns, from [134]. cial Semantic Web is still needed. Premises and conclusions may also be chained (as in (iv) of Figure 1 on the next page). Further, as we will see in this review, there are differ- 3. Requirements ent ways of thinking about and modeling arguments. Argumentation theorists have variously modeled indi- What are the requirements for supporting argu- vidual argument structure (e.g. Toulmin, discussed in mentation on the Social Semantic Web? Arguments Section 4.1, page 4), argument chaining (e.g. Arau- must be identified, resolved, represented and stored, caria3 [150,142,144]), and groups of arguments (e.g. queried, and presented to users. Identification involves Dung, discussed in Section 4.5, page 6). There is a sig- mining arguments, in the form of claims, from text nificant difference between what is possible to analyze (Section 6.12.2, page 17), eliciting them from users, when looking at these different levels of argumenta- or some combination of these approaches.