Soft Systems Methodology (SSM) as a Viable Methodology for Knowledge Engineering: A Literature Review (Research Paper Sample)

Soft Systems Methodology (SSM) as a Viable Methodology for Knowledge Engineering: A Literature Review
Chukwunonso, F.*, Ibrahim, R.* and Selamat, A.*
* Department of Information Systems, Faculty of Computing, Universiti Teknologi Malaysia, 81310 Skudai, Malaysia
Article Info


ABSTRACT

Article history:
Received Jun 12th, 201x
Revised Aug 20th, 201x
Accepted Aug 26th, 201x


This paper presents a baseline perspective of knowledge engineering (KE) methodologies by taking a critical look at the methodological approaches currently used in knowledge engineering domain. A literature review bordering on; the knowledge role concept, knowledge elicitation techniques, and ontological approaches employed in knowledge engineering for the construction of knowledge-based systems (KBSs) was carried out. Journal and conference articles were sourced from multiple journals and research databases and a multi-step manual cross-checking based on carefully selected extraction and quality criteria were employed. The findings of the study show that the existing methodological approaches employed for constructing KBSs in KE is highly deficient and inefficient for solving KE problems under dynamic and uncertain environments. The paper concludes by presenting a strong argument as to why soft systems methodology is best suited for constructing knowledge base systems (KBS) in a spatially distributed, unstructured and shared domain specific context.

Keyword:
Knowledge-Based Systems
Knowledge Engineering
Expert Systems
Soft Systems Methodology





Copyright © 201x Institute of Advanced Engineering and Science. All rights reserved.

Corresponding Author:
Franklyn Chukwunonso,
Department of Information Systems,
Faculty of Computing,
Universiti Teknologi Malaysia, 81310 Skudai, Johor, Malaysia.
Email: fchukwunonso2@live.utm.my

1 INTRODUCTION
Knowledge engineering originated in the 1970s out of expert systems, following the need for a systematic method to build knowledge-based systems (as it was referred to in the early days). [1] first defined knowledge engineering as the process of reducing knowledge from a large body into a set of precise rules and facts. This definition was extended by [2] to include the necessity for gaining more understanding of the characteristics of expertise in itself as well as understand how this knowledge can be applied in eliciting expert knowledge in domain-specific contexts. As noted by [3], the early application of expert systems in knowledge engineering had a lot of setbacks due to its unstructured nature. Several attempts have been made in the past as to how to overcome this bottleneck especially in the area of knowledge elicitation. The initial purpose of knowledge elicitation was aimed at transferring expertise knowledge to the knowledge engineer in the field of software development [4]. Though some of the results of previous researches looked promising, the challenge of transferring technology from AI to developing KBSs proved abortive. [5] attributed these failures largely to small natures of KBSs developed thereby making it difficult to explore the feasibility of different methodological approaches. The complexity of the problem is directly connected to the challenge faced in the late 1960s during the "software crisis”, when the methods used for constructing traditional software system prototypes could not scale up to designing and maintaining large and sustainable commercial software for solving real life problems, which later led to establishing Software Engineering as a discipline. In the same vein, KE discipline was established for the primary purpose of transforming the processes involved in KBS’s construction from an art to an engineering discipline in order to create a better analysis and understanding of the processes involved in building and maintaining KBSs, and to develop suitable methods, specialized tools and languages for the construction of KBSs. The rest of the paper is therefore structured as follows: Section 2 takes a critical look at the concept of knowledge role in KE. Section 3 briefly discusses the selection criteria methods employed for the study. Section 4 discusses findings of the literature review and Section 5 concludes the study, stating its limitations and areas of further studies.
2 STUDY BACKGROUND
In the early days, expert systems were used to separate domain knowledge from general reason to form sets of knowledge base rules. In the early 1980s, several studies identified setbacks to this approach and proposed the use of a systematic approach to KE. [6] noted that there are different levels at which knowledge is attained and that it should be considered when solving knowledge-based tasks. He posits that the knowledge level, which is higher than the symbolic level, addresses issues regarding knowledge representation such as why a system or agent performs an action independent from logic, symbolic frames or rules (symbolic level of knowledge). Currently, these description used by [6] at the knowledge level has since been the basic principles on which knowledge engineering is founded, and has provoked several other studies such as [7] in this regard. [8], [9] and [10] distinguished between the different types of knowledge in a knowledge-based system and in a related study, [11] provided a heuristic classification of the standard patterns used in solving knowledge-level problems [12]. Their findings later became very useful in solving knowledge engineering problems related to task knowledge. The 1990s saw a shift in focus from task knowledge to domain knowledge. Knowledge representation took center stage with ontology approaches becoming widely used for representing knowledge [13-14]. The increased use of ontologies for concept sharing in a distributed knowledge domain like the World Wide Web (WWW) led to the development of several modelling languages including Ontology Web Language (OWL) presently used in semantic web. Also, these modelling domain languages make use of patterns like in task knowledge.
Knowledge role can be simply defined as the role a particular knowledge domain plays in solving a particular problem [9]. This helps in structuring the problem by imposing constraints on the way a particular knowledge domain may be used in the course of reasoning, thereby increasing the feasibility of the problem solving process, unlike the uniform reasoning method used in the traditional expert systems where one large knowledge base applies to all [15]. Some examples of knowledge role commonly used in assessment method include decision, norm, and case data. [16], in his review of KE techniques, recommended [17] Personal Construct Psychology (PCP), automated by [18] and modified by [19]. PCP presents a model that addresses the unstructured nature of human psychology in representing, acquiring, and processing of knowledge. He argues that PEGASUS, a computer program developed by [20-21] were more suitable for encoding aspect of human reasoning based on the vocabulary of experts into formal concepts and structured knowledge. [22] also asserted to this by suggesting it could also be used in teaching by allowing other teachers make use of one expert’s vocabulary in same way as the expert (teacher in this case). The major setback with this technique though is that it bridges the principles of psychology used in PCP with its logics and systemic principles when producing a framework for KE. Also, [23] noted that verbal reports had no correlation with mental behavior in his study on psychologist’s attitude towards verbal data from patients. Furthermore, the hierarchy employed by PCP assumes a strongly formal and idiosynchratic postulation based on uniformity. This is wrong in the notion of soft systems, which assumes that there is no valid "right answer" for all situations, in other words, there is no one-size-fits-all approach.
3 METHODOLOGY
The literature review was based on the structure provided by [24]. The study adopted the guidelines stated by [25]. The review was framed by the questions which this research tried to answer and influenced the search criteria employed for the study. The research questions covered five different subdomains of KE based on the scope of and objective of this study which bordered on the knowledge role concept, knowledge elicitation techniques, and ontological approaches employed in knowledge engineering for the construction of knowledge-based systems (KBSs). The research questions are: a) What are the current knowledge elicitation techniques employed in KE? b) What are the current methodological approaches used in KE? c) What role does Ontology play in KE? d) What special features distinguishes KE from other disciplines? and e) Why is SSM considered as the most suitable and efficient method for KE problem-solving in a spatially distributed and shared domain specific context? In conducting this literature review, both journal and conference articles were sourced manually from several online databases and selected based on the criteria stated above, through a multi-step manual filtering process with independent validation at each step. Duplications and overlapping of selected papers were manually sorted out. The use of manual processes ensured rigor, relevance and credibility was maintained during the selection process. The full texts of selected papers were then independently analyzed by the researchers to ensure it meets the quality assessment criteria presented in Table 1, for inclusion. This was also used to validate the articles selected for this review study.
Table 1. Quality Criteria [25]
1

Is the paper based on research (or is it merely a ‘‘lessons learned" report based on expert opinion)?

2

Is there a clear statement of...