Learned social ontologies can be viewed as products of a social fermentation process, i.e. a process between users who belong in communities of common interests (CoI), in open, collaborative, and communicative environments. In such a setting, social fermentation ensures the automatic encapsulation of agreement and trust of shared knowledge that participating stakeholders provide during an ontology learning task. This chapter discusses the requirements for the automated learning of social ontologies and presents a working method and results of preliminary work. Furthermore, due to its importance for the exploitation of the learned ontologies, it introduces a model for representing the interlinking of agreement, trust and the learned domain conceptualizations that are extracted from social content. The motivation behind this work is an effort towards supporting the design of methods for learning ontologies from social content i.e. methods that aim to learn not only domain conceptualizations but also the degree that agents (software and human) may trust these conceptualizations or not.
TopIntroduction
Although Semantic Web has lately moved closer to its pragmatic realization, a critical mass of semantic content is still missing. Web users can only find few well-maintained and up-to-date domain ontologies. Only a small number of Web users, typically members of the Semantic Web community, build and publish ontologies. To assist and motivate humans in becoming part of the Semantic Web movement and contribute their knowledge and time to create or refine/enrich useful ontologies, there is need to advance semantic content creation by providing Web users with a “starting point of assistance” i.e. with automatically learned kick-off (starting-up) ontologies.
Traditionally, the learning of ontologies involves the identification of domain-specific conceptualizations that are extracted from text documents or other semi-structured information sources e.g. lexicons or thesauruses. Such learned ontologies do not utilize any available social data that may be related to the domain one e.g. information ownership details (contributor, annotator or end-user), tags or argumentation/dialogue items that have been used to comment, organize or disambiguate domain-specific information, querying information related to user clicks on retrieved information. Recently, the learning of ontologies also concerns social content that is mainly generated within Web 2.0 applications. Social content refers to various kinds of media content, publicly available, that are produced by Web users in a collaborative and communicative manner. Such content is associated to social data that has been produced as a result of the social fermentation process. The most popular social data in Web 2.0 content is tags, which are often single words listed alphabetically and displayed with a different font size or color to capture its importance. Tags are usually hyperlinks that lead to a collection of items that each tag is manually or automatically associated with. Such social data can be processed in an intelligent way towards shaping social content into ontologies. Since social data is created as part of the social fermentation process, that is, tags are introduced in a collaborative and communicative manner; it can be argued that the learned ontologies produced from such a process encapsulate some degree of agreement and trust of the learned conceptualizations.
Social content generation (SCG) refers to a conversational, distributed mode of content generation, dissemination, and communication among communities of common interest (CoI). Social intelligence (SI) aims to derive information from social content in context-rich application settings and provide solution frameworks for applications that can benefit from the “wisdom of crowds” through the Web. Within this setting, a social ontology can be defined as: an explicit, formal and commonly agreed representation of knowledge that is derived from both domain-specific and social data. In the context of this chapter, the meaning of the term “social ontology” must be clearly distinguished from the meaning that is used in social sciences. A representative definition in the context of social sciences is given by T. Lawson of the Cambridge Social Ontology Group1: “…the study of what is, or what exists, in the social domain; the study of social entities or social things; and the study of what all the social entities or things that are have in common”.
Formally, an ontology is considered to be a pair O=(S, A), where S is the ontological signature describing the vocabulary (i.e. the terms that lexicalize concepts and relations between concepts) and A is a set of ontological axioms, restricting the intended interpretations of the terms included in the signature (Kalfoglou & Schorlemmer, 2003; Kotis et al, 2006). In other words, A includes the formal definitions of concepts and relations that are lexicalized by natural language terms in S. In this chapter, we extend such model by adding a social dimension (equal to social semantics) that is inspired by the definition of the “Actor-Concept-Instance model of ontologies” (Mika, 2007) formulated as a generic abstract model of semantic-social networks. This extended model is build on an implicit realization of emergent semantics, i.e. meaning must be depended on a community of agents. According to the extended model, a social ontology is defined as a triple O=(C, S, A), where C is the set of collaborated contributors that have participated in a SCG task, from which S and A have been derived using the SI found in C. The range however of C over both S and A at the same time is not guaranteed, i.e. S may have been derived from C, but not A, which may have been automatically derived from external information sources such as a general ontology or lexicon, e.g. from WordNet.