I1-[OntoSpace] during first phase (2003-2007)
Summary:Natural language is an essential mode of interaction between users and sophisticated spatially-aware systems such as mobile assistance robots. Providing suitably sophisticated natural language capabilities for ever more complex interaction scenarios is a major problem.
An important contributing factor is the lack of appropriate modularizations of the technical components involved in complete systems that combine spatial and linguistic capabilities. Such interaction needs to be as natural and non-intrusive as possible in order to support the widest possible range of potential users, and so sophisticated accounts are required. Most of the features that make interaction natural still present substantial challenges for dialog systems and solutions are not to be found within single research areas. One particularly effective strategy for modularization is the adoption of linguistically motivated ontologies that mediate between domain/application knowledge and Human Language Technology (HLT) components. But current strategies for ontology mediation exhibit a rigidity that is inappropriate for the relationships observed between domain and linguistic knowledge in real interactions. Here, a negotiation of mediation within conversational interaction appears crucial both for ontology design and for achieving inter-operability.
This project therefore has the primary goal of developing a toolbox of ontology-based methods suitable for supporting natural language interaction and technology re-use within a range of interaction scenarios involving spatially-aware systems. The methods developed will be motivated by detailed empirical investigations of the actual communicative strategies of negotiation employed by users of robotic systems. The results of these experiments will feed into the formalization of inter-ontology mappings. The project will contribute to the further development of emerging standards in ontological engineering, their application in the management of natural language, and the development of ontological modules involving spatial representations for mobile robots. It will also use the ontology mediation strategy to overcome a persistent lack of interaction between spatial system design, robotic interaction, etc. and HLT. This has limited both the wider re-usability of important research results concerning the natural language analysis and generation of spatial relations, including fine-grained semantics for spatial expressions, route description planning and understanding, resource-adaptive generation, etc.
|SFB/TR 8 Spatial Cognition|