Matchmaking augmented with general Knowledge

Description Logics (DLs) are well suited for the representation of terminological knowledge and support efficient reasoning about these terminologies. In this use case a large set of classified advertisements (ads) has to be matched with queries, derived from individual readers search requests as well as management needs to track the amount of classified ads in certain areas of interest. Queries and ads are represented by concept expressions Cq and Cad . An ad is a match of a given query iff Cad v Cq with respect to the domain knowledge represented in the TBox, containing concepts and concept inclusions of the description logic knowledge base. For every advertisement a concept expression is generated from the highly abbreviated text. Only the most necessary information is printed to save space and expenses. In addition to these economic reasons the reader shall not be bothered by redundant information. As a consequence only those features of goods are mentioned in ads, that are not commonly assumed. Caused by this restricted text the retrieval of the matching was weak. The common knowledge has to be included in the process of matching i.e. the subsumption has to recognize general knowledge and expectations that are not mentioned in Cad . This inclusion has to be compatible with exceptions from the general. To integrate this general knowledge with the matching process the TBox T is split into the standard part TS and the general or default part TD. The general knowledge has to be applied, iff the subsumption is uncertain wrt. TS. A subsumption C v D is called uncertain C 6vTSD with respect to a TBox TS, if neither C v D nor C v D are a consequence of TS. Based on the split TBox and uncertainty the default subsumption can be defined. The default subsumptionCvT D shall hold iff it is a consequence of TS or in case of uncertainty a consequence of TS∪TD for a satisfiable concept C. T `C v D ⇐⇒ TS `C v D ∨ ((C 6vTSD)∧ (TS∪TD `C v D)∧ (TS∪TD 6`C v⊥))