Semantic modelling of product support knowledge
knowledge model, which facilitates the sharing and reuse of product support knowledge. The applicability of the approach is illustrated using examples extracted from the knowledge based system developed.
Thank you for your comments and questions.
It seems that you got confused with the examples of the predicates describing how the notions “knowledge-specifier” etc. can be applied. The knowledge specifier itself is “Product is complex” for example. Apparently this cannot be expressed with an “if-else” rule, which denotes action. The advantage of distinguishing knowledge from action lies in the fact that knowledge once formally defined and by being machine processable (ontologies deliver both of the aforementioned qualities), can be not only easily reused (into defining several different actions by reusing the same pieces of knowledge in different ways) but also can be easily understood (avoiding redundancies and confusions) by several different people and functional units.
Product support modelling is unified by defining the structural notions required for expressing that knowledge, and by defining what each of these notions represent. Different knowledge-specifiers can be added by a knowledge engineer, for example, in different occasions, although after careful consideration and studying of several previous cases in product support some basic properties such as “Product is complex” have been found to be always needed in product support.
Ontological principles include ontological axioms (such as definitions of what is a supercalss, a concept, etc.), guidelines of development methodologies (e.g. METHONTOLOGY) and other philosophical based principles such as unity, identity, essence, and dependence.
An example of adaptive product support has been given with the predicates (novice user takes no support for complex product while experienced is provided with support). An example of personalized support was not given but by having different user classes, different actions can be carried out for each different class.
Previous work is adaptive but the adaptation process is linked only to the user’s characterization without considering the product or the task under question, which is provided in this paper. As an example previous approaches were considering the characteristics of a runner (e.g. age, stamina) without taking in mind the characteristics of the race (e.g. in a stadium or on different surfaces in the countryside) or the kind of the race (e.g. cycling and running instead of only running for example).
Quantifiable comparisons have been provided for the use of ontologies in knowledge engineering in the literature, which prove that by reusing knowledge and facilitating interoperability (two basic functions of ontologies) significant saves in costs have been achieved, while the development time has been reduced.
Hi,I have several questions about this paper, as I am not clear enough about some of the opinions in the paper. Many thanks!
1. About "knowledge-specifier", "arc", "connector-dependece" measure.
From the explanation in the paper, I understand they are "if-thens", and I think all of them can be represented as "if-thens" as rules. so how do you decide which is "considered so significant within the application domain"?
How is it different from the form of "if-thens", any benefits?
If "products support modelling is unified", so why "that is considered so sigificant with the application domain", does it mean they are application dependent?
If yes, then how do you "unify" them? 2."ontological principles", you mentioned in the paper, so what are they?
3."it 'facilitates' the provision of adaptive and personalised support" from the conclusion. As they are not stated clearly in the paper how it can achieve this. Can previous work achieve this?
4. From the references, I can see "Adaptive product manuals", so I think previously they can achieve this "adaptive" goal, so I am interested to know how it is compared with the solution in this paper.
Any benefit? I mean from cost, developing time, etc. I think there must be some benefits over previously proposed methods from some aspects, so if there are some quantitive measurements, it will be more convincing.
All in all, interesting paper. Thank you!