Semantic Matchmaking of Web Resources with Local Closed-World Reasoning

Stephan Grimm, Pascal Hitzler
2007 International Journal of Electronic Commerce  
Ontology languages like OWL allow for semantically rich annotation of resources, such as products advertised at an electronic online marketplace, while the Description Logic (DL) formalism underlying OWL provides reasoning techniques to perform matchmaking on such annotations. We identify peculiarities in the use of DL inferences for matchmaking which are due to the open-world semantics of OWL, and we analyse the use of local closed-world reasoning for its applicability to matchmaking. In
more » ... ular, we investigate two nonmonotonic extensions to DL, namely autoepistemic DLs and DLs with circumscription, for their suitability of realising local closed-world reasoning in the matchmaking context to overcome these problems. We discuss their different characteristics by means of an elaborate example of an electronic marketplace for PC product catalogues from the eCommerce domain and demonstrate how these formalisms can be used to realise such scenarios. 2004. OWL is essentially based on description logic (DL) [3], a logic-based formalism with well understood computational and representational properties. It thus allows for logical reasoning, and matchmaking over OWL-DL can be operationalised by techniques of automated deduction. As a fragment of first-order predicate logic, OWL inherits a semantics that adheres to the open-world assumption (OWA), under which situations of incomplete information can be handled by the distinction between negative knowledge and the absence of knowledge. At the same time, it also inherits features which make it awkward for representing certain kinds of knowledge. In particular, OWL does not allow for any form of closed-world or default reasoning in such situations. The quest for alterations or extensions of OWL to make it more suitable for important modelling tasks is currently one of the prominent research issues in ontology language research. While such investigations are frequently pursued on the basis of formal logical argumentation, the natural way of deriving requirements from application needs is often neglected. In this paper, we therefore investigate the usability of OWL for the modelling of matchmaking problems. In particular, we will point out that a pure open-world semantics leads to unintuitive behaviour in certain cases. Going one step further, we will study recent approaches to weakening the open-world assumption in OWL by means of local closed-world features [13] , and examine them with respect to the matchmaking requirements we have identified. This paper thus serves different purposes. It identifies problems due to the open-world assumption in an elaborate discussion of an example taken from the eCommerce domain. It elaborates on the proposal of using a local closed-world semantics to overcome these problems, based on the ideas in [18] . It presents an application of two particular nonmonotonic extensions to DL, namely autoepistemic DL and DL with circumscription, to the problem of matchmaking with respect to their suitability for realising local closed-world reasoning in this context. The detailed example demonstrates how the constructs of these formalisms can be applied to realise matchmaking in a particular scenario. The overall presentation is based on arguments at an intuitive level, however, the interested reader will also find the necessary formal details required to map the intuitive argumentation to the underlying details of the logics used. Readers not primarily interested in the model-theoretic realisation of matchmaking can skip these technicalities without missing the main messages conveyed. The paper is structured as follows. In Section 2 we will briefly introduce OWL and approaches to local closed-world reasoning with the DL underlying OWL, namely autoepistemic description logics and description logics with circumscription. In Section 3 we will show how resources for matchmaking are modelled in OWL, and in particular we will introduce a large running example which will help us to illustrate our discussion. In Section 4 we show how DL inferencing is used for matching resource descriptions, and we will identify deficiencies of modelling based on the open world assumption. The forms of local closed-world reasoning introduced in Section 2 will then in Section 5 be analysed as to their ability to rectify the identified problems. Section 6 discusses related work, and in Section 7 we will summarise and point to concrete research issues which need to be addressed in the future.
doi:10.2753/jec1086-4415120204 fatcat:jwanagqouvg35bio5vfctl4dua