A social software/Web 2.0 approach to collaborative knowledge engineering

Abstract

Expert systems have traditionally captured the explicit knowledge of a single expert or source of expertise in order to automatically provide conclusions or classifications within a narrow problem domain. This is in stark contrast to social software which enables knowledge communities to share implicit knowledge of a more practical or experiential nature to inform individuals and groups to arrive at their own conclusions. Specialists are often needed to elicit and encode the knowledge in the case of expert systems, whereas one of the (claimed) hallmarks of social software and the Web 2.0 trend, such as Wikis and Blogs, is that everyone, anywhere can chose to contribute input. This openness in authoring and sharing content, however, tends to produce unstructured knowledge that is difficult to execute, reason over or automatically validate. This also poses limitations for its reuse. To facilitate the capture of knowledge-in-action which spans both explicit and tacit knowledge types, a knowledge engineering approach which offers Wiki-style collaboration is introduced. The approach extends a combined rule and case-based knowledge acquisition technique known as Multiple Classification Ripple Down Rules to allow multiple users to collaboratively view, define and refine a knowledge base over time and space.

Introduction

Sharing what we hold in our heads is a major challenge facing postindustrial knowledge economies. Some of what we “know” can be articulated, codified, structured and stored in repositories such as books, databases and knowledge bases. In the past the focus of computer systems has been on managing data, information and explicit type knowledge, such as was found in expert systems. More recently techniques have emerged for acquiring softer (or implicit) knowledge using social software such as Wikis, online Communities of Practice (CoPs) and WebLogs which also “helps to realize the original vision of the Web as a space where anyone can participate” [19, p. 2]. These emergent behaviours reveal a trend, known as Web 2.0, for creative uses by individuals of existing internet-based technology involving information sharing and collaboration. Recognising that research has shown that decisions are arrived at through the combined use of almost equal amounts of tacit and explicit knowledge [10], we seek to leverage the collective benefits of formalised knowledge in traditional expert systems and informal knowledge such as we find in Web 2.0 applications. We¹ propose an approach which uses problem situations as they arise to motivate the capture and description of the problem and the explicit rules which define how to handle such incidents. In this way we bring together expert systems which provide structured and reusable knowledge, often in the form of production rules, with less formal knowledge shared amongst social networks of people, often in the form of cases, experiences and stories.

Traditional approaches to knowledge engineering, as with use of the traditional Waterfall model in software engineering, may be appropriate when the goal is to build a static model of a narrow, well defined and well-understood domain but they do not work well for dynamic domains and evolving knowledge [9]. While the resulting knowledge model may have been shared, it was not engineered through incremental acquisition and refinement of knowledge by multiple distributed users with different levels of expertise and conflicting views and knowledge. In summary, “When this [traditional] process [is] applied to a more flexible domain like social networks and knowledge communities, it lacks the opportunity of changing and redefining the chosen knowledge structures… Often the classical knowledge engineering process in social networks leads to the contra productive result that the built knowledge structures, their semantics and meaning are challenged … The problem of maintaining and adapting knowledge models to flexible needs of the users takes a lot of effort” [11, p. 166].

In knowledge engineering, ontologies are typically offered as the means to support development of a shared understanding. The irony is that most approaches to ontology development do not support collaborative ontological engineering as shown in a comparative study of ontological engineering tools [5]. Ontolingua was an exception and allowed editing by multiple users but other users were not notified of changes and who had done them. More recently Bao and Honavar [1] conducted a similar study and found that while many of the ontology editors provide concurrent access control with transaction oriented locking and in some cases even rollback, “there is an urgent need for principled approaches and flexible tools for allowing individuals to collaboratively build, refine, and integrate existing ontologies as needed in specific contexts or for specific applications” [1, p. 8].

Consistent with our experience, the Software Engineering Body of Knowledge (SWEBOK) project [11] found that in order to support social networks of people to collaboratively engineer a knowledge base it is necessary to:

• build a shared repository which achieves a common understanding and structure together with a means of finding out what projects or problems people are working on (providing opportunities for sharing and reuse);

• allow people to create content in their own way using their own terms and concepts;

• develop a top-down knowledge map in the form of an ontology or concept map to assist people to define and structure their own concepts within the bigger picture;

• allow the knowledge nuggets to emerge bottom-up, or top-down where appropriate, in an ad hoc, immediate and spontaneous manner.

  Additionally, we found the system needs to:

• support a range of expertise levels, views of the knowledge and access rights;

• provide a review process in which users can register their approval or disagreement;

• test and keep track of the consistency between all elements of the knowledge system and notify users when conflicts occur;

• be compatible with a wide range of existing systems and knowledge sources;

• provide an intuitive, simple and yet structured knowledge maintenance cycle;

• support primary ownership and management of the domain knowledge by domain users, not a third party such as a knowledge engineer.

  Finally, a collaborative knowledge management tool needs to:

• be configurable to each enterprise allowing changes to the ontology, the workflow/process of acquiring knowledge and the user interface over the life of the system [26].

This paper presents a system design and prototype which embodies the above 11 requirements. In the next section the motivation for collaboratively building knowledge based systems is discussed further. Section 3 describes some approaches to collaboration. Section 4 introduces the approach known as C-MCRDR. Contributions and conclusions are found in Section 5.