Multi-aspectual knowledge elicitation
Winfield, MJ 2000, Multi-aspectual knowledge elicitation , PhD thesis, Salford : University of Salford.
Restricted to Repository staff only until 31 January 2017.
Download (5MB) | Request a copy
This work examines one of the major stumbling blocks of knowledge based systems development, namely knowledge elicitation. The challenge is a fundamental one of eliciting knowledge from domain experts including tacit knowledge. This thesis argues that, in the past, knowledge elicitation has been limited since elicitation has been performed from one or a limited number of aspects. A method is needed to assist in providing a pluralistic approach to knowledge elicitation that will aid multi-aspectual viewpoints of the domain knowledge to be elicited. MAKE (Multi-Aspectual Knowledge Elicitation) is such a pluralistic method. Using the work of Herman Dooyeweerd (1955) MAKE is developed from a sound philosophical basis. Two levels of knowledge are elicited using MAKE. The method starts by building a top-level knowledge map that covers all of the knowledge aspects and provides an overview of the domain. Such an overview determines the complexity of the domain allowing a knowledge based systems developer to see the effects of taking a minimalist approach to the development; that is the top-level map may be used to help define the scope of a system. The second level involves detailed knowledge elicitation. Using a process of abstraction, the concepts defined in each aspect are refined to a sufficiently detailed level to enable a system to be built. The resulting knowledge forms an ontological view of the domain knowledge. The empirical work adopting a case study approach has demonstrated that: MAKE can be used by people who are not necessarily versed in artificial intelligence techniques or in the philosophy of Dooyeweerd. MAKE has shown itself to be adaptable across a very varied set of domains MAKE is adaptable and useful for eliciting tacit knowledge. It is argued that MAKE indicates a change of direction from methods that are currently in use.
|Item Type:||Thesis (PhD)|
|Contributors:||Basden, A (Supervisor)|
|Schools:||Schools > School of Computing, Science and Engineering|
|Depositing User:||Institutional Repository|
|Date Deposited:||03 Oct 2012 13:34|
|Last Modified:||01 Dec 2015 00:04|
Actions (login required)
|Edit record (repository staff only)|