Particularly in computational approaches to natural language processing and intelligent systems, the issue of relating linguistic ontologies to particular 'domain' ontologies is an important one - the organisational motivation of the two kinds of ontology are often quite different and a mapping between them is not straightforward. One such tradition is that of the 'Upper Model' (Bateman/Kasper/Moore/Whitney, 1989; Bateman/Henschel/Rinaldi, 1995: http://purl.org/net/gum2), a linguistically motivated event-based ontology used in natural language generation where many questions have been raised about how best to interface with knowledge of particular domains.
Whereas the Upper Model in its CONCEPTION shares certain features with more recent developments in linguistic ontologies - such as the EuroWordNet top ontology, DOLCE, and others - we have been led in its USE to a recognition of issues so far backgrounded in ontological investigations: in particular, we are now focusing on the very active role played by linguistic INTERACTION in establishing mappings between linguistic and non-linguistic configurations. Results in psychology (e.g., Garrod/Sanford, 1988) have long drawn attention to the mediating role of natural language interaction in setting up correlations between the interpretation of linguistic expressions and their possible referents in the world. We are now investigating whether this can be seen as a process of guided ontological mediation employing methods from ontological engineering.
We believe that several theoretically important questions can be addressed from this perspective. In particular, it is common to overlook the variability that exists between linguistic forms and their non-linguistic ontological correlates. It appears in natural interactions that this variability is constrained and negotiated by the interaction partners. Examining the range of variation that is open to negotiation will give us a firmer hold on the structure and contents of the individual levels that are being related. Such information should substantially supplement constraints being investigated from other perspectives.
Our experimental setting for the investigation is drawn from the area of spatial reasoning and communication. By setting up constrained tasks involving spatial configurations and movement that are to be achieved cooperatively by human and artificial systems (such as robots or intelligent wheelchairs), we are investigating the flexibility required in mapping between the Upper Model linguistic ontology used for natural language generation and analysis and domain ontologies developed for spatial representation (Moratz/Tenbrink/Bateman/Fischer, in print).
Bateman, J.A., R.T. Kasper, J.D. Moore, & R.A. Whitney (1990) A general organization of knowledge for natural language processing: the PENMAN upper model. Technical report, USC/Information Sciences Institute, Marina del Rey, California, 1990.
Garrod, S.C., & A.J. Sanford (1988) Discourse models as interfaces between language and the spatial world. Journal of Semantics, 6:147-160.
Moratz, R., T. Tenbrink, J. Bateman, & K. Fischer. (in print). Spatial Knowledge Representation for Human-Robot Interaction. In Freksa, C., Brauer, W., Habel., C. & Wender, K.F. (eds.), Spatial Cognition III. Berlin, Heidelberg: Springer
| Workshop Homepage | Workshop Program | Next Abstract |