Knowledge Engineering Rediscovered: Towards Reasoning Patterns for the Semantic Web

Abstract

The extensive work on Knowledge Engineering in the 1990s has resulted in a systematic analysis of task-types, and the corresponding problem solving methods that can be deployed for different types of tasks. That analysis was the basis for a sound and widely accepted methodology for building knowledge-based systems, and has made it is possible to build libraries of reusable models, methods and code.In this paper, we make a first attempt at a similar analysis for Semantic Web applications. We will show that it is possible to identify a relatively small number of tasktypes, and that, somewhat surprisingly, a large set of Semantic Web applications can be described in this typology. Secondly, we show that it is possible to decompose these task-types into a small number of primitive ("atomic") inference steps. We give semi-formal definitions for both the task-types and the primitive inference steps that we identify. We substantiate our claim that our task-types are sufficient to cover the vast majority of Semantic Web applications by showing that all entries of the Semantic Web Challenges of the last 3 years can be classified in these task-types.

Enhancing Content-Based Recommendation with the Task Model of Classification

“…For the first two tasks (browse and search), we adopted the definitions from van Harmelen and ten Teije [9]. In this paper, we focus on defining the third task (content-based recommendation) and analyzing the corresponding inference steps.…”

“…The standard content-based recommendation (CBR) usually takes the user profile plus the domain ontology and returns a set of instances, which might be of interest to the user [9]. In the case of CHIP, the system stores the user profile in the form of both a set of concepts and a set of instances.…”

“…Compared with the original definition of recommendation and its corresponding inference steps from van Harmelen and ten Teije [9], we mainly extended the inference step of classification, which now consists of two components: classification by concepts and classification by instances. The original classification only determines where a given class should be placed in a subsumption hierarchy.…”

“…However, most work deals with reusable ontology and ontology patterns [1], there is hardly any work on reusable reasoning patterns, except the work from van Harmelen and ten Teije [9]. They made a first attempt at finding reusable task types and decomposing these tasks into a number of primitive reasoning patterns for Semantic Web applications.…”

Towards a Theoretical Foundation for the Harmonization of Linked Data

“…Other approaches include the formalization of prototypical tasks through reasoning patterns [6], approaching the semantic web as a unique ontology and not as linked data, and the field of ontology matching [7], which focuses on the ontological level (the OWL level) instead of the data structure (the RDF level). Existing approaches however do not deal with the problem of specifying data requirements and ensuring reliability with respect to these requirements.…”