Overall strategy and general description
The above use case studies illustrating life with and without OASIS all rely on the same essential move beyond what is currently possible: particular services can rely on information that is relevant for their activities and assistance but which was not previously readily available to them in any form that could be used. In any particular case, cost-intensive connections could be built, but this in reality simply reduces the chances that the people who need the services, in the OASIS case the elderly, would ever see them realised. The approach of OASIS is therefore simple in conception: direct re-usability of information is to be provided across heterogeneous services and devices. Much of the individual components of such advanced re-usability are already in place. Various service providers currently provide or are interested in providing spatial-related assistive services for the elderly, including services in the fields of domotics, travel and leisure, transportation, etc. Much of the complexity that arises for providing intelligent assistance then revolves around issues of effective and efficient communication between the user and their assisting device, as well as communication between different elements of the assistive service (e.g. assistive device and sensors). In all cases, to achieve efficient assistance services, it is necessary to ensure a ‘common understanding’ of contextual information about different services and objects. Such common understanding and sharing of contextual information can only be achieved by enabling the interconnection of heterogeneous data models used by each service (i.e. by each service provider) or even module of service. This will in turn enable the integration of the plethora of diversiform services into a common platform in order to provide not only improved but also new services for the elderly.
In order to achieve interoperability of services and sharing of contextual information between different services and objects, it is necessary to model them first, by extracting each service’s individual structure up to its most primitive level. In current approaches, this can lead to more or less ad hoc solutions. The OASIS solution is to provide foundational ontology components, specifically tailored to the requirements of the applications to be covered and the services provided. In addition, for new components, ontology design principles as well as guidelines for best practice will be produced to make modelling OASIS-compliant from the outset. Moreover, the selection of appropriate ontological components for service applications is currently further hindered by the fact that it is virtually impossible for a developer to accurately estimate the costs involved in advance. The OASIS method will provide an effective characterisation of the foundational components that are available and a checklist that relates ontological complexity (including both content and reasoning requirements) to the concrete application demands to be met. This will provide the basis for a far more accurate in advance prediction of involved costs that also goes beyond what is currently available.
Traditional uses of ontology generally require the alignment of domain ontologies against some high-level model defined in advance. The OASIS approach sees this as only one of several solutions: an OASIS-compliant ontology may be aligned in the OASIS hyper-ontological framework as common practice suggests. A further, more flexible solution is for the domain ontology to be related to the OASIS hyper-ontological framework by just those inter-ontology mappings necessary for inter-operability. This approach does not require complete alignment and also supports more complex inter-ontology relationships. Again, ease of mediation will be maximised by building up a library of established inter-ontology relations drawing on modular foundational components as demanded by the real applications covered in the demonstration scenarios. This means that any application can call any service without knowing its structure a priori, but only by being aware of the common data delivery format defined in the hyper-ontology.
The OASIS foundational framework will therefore provide an open and extensible hyper-ontology (Common-Ontological Framework: COF) containing two major resources: (i) a library of component ontologies and ontological modules supporting inter-operability across the knowledge sources and services of the application scenarios, and (ii) a library of inter-ontology mappings maintained within a formalised, logically well-founded generic framework that supports the addition of new ontologies and the construction of complex services drawing seamlessly on diverse information sources. An further essential feature of the OASIS approach will be to support diversity not only in terms of the types of information modelled but also in terms of the complexity of that information: the hyper-ontology will enforce a modularity that allows both highly expressive (but computationally expensive) ontologies to exist alongside straightforward and computationally inexpensive ontologies such as those most commonly applied in practical applications. The hyper-ontology mappings between ontologies of different complexity is a major innovation that will support the gradual and controlled migration of capabilities supported by complex foundational ontologies into practical contexts of use.
OASIS will develop these new techniques by adapting for the ontological engineering task tools that have already proved themselves in the software specification community. In particular, we adopt the de facto standard for software specification CASL (Common Algebraic Specification Language) produced as part of the Common Framework Initiative (COFI) and approved by IFIP WG 1.3 “Foundations of Systems Specifications” [Astesiano02]. Within CASL, issues of heterogeneity (diverse logics), relations between logics (institutions and category theory), as well as large-scale application for realistic applications have all been addressed and receive substantial treatment. A development environment for CASL, called the Heterogeneous Tool Set (HETS), is under continual development at the University of Bremen and will be employed as one of the formal foundations for the OASIS Common Ontology Framework. Both theoretical and practical experiences gained within the OASIS development will feed into both CASL and HETS. In prior work we have demonstrated that CASL is a serious contender for a new standard for specifying heterogeneous modularised ontologies [Lüttich04]. OASIS will build on these results, defining ontologies and inter-ontology mappings that refine and feedback into the basic theory as shown necessary, during cycles of application evaluation in practice.
Ontologies developed within CASL are highly modular and are formally axiomatised within any of the logics supported by CASL. These ontologies are automatically linked via HETS to logic-specific reasoners. Thus, a module that is expressed within a limited logic, such as OWL-DL, can be used with an efficient description logic reasoner such as Racer, whereas a module requiring a more expressive logic (such as Wonderweb’s DOLCE ontology [Masolo03] or the Suggested Merged Upper Ontology, SUMO: [Pease02]), are used with full first order reasoners, such as SPASS. A major current research issue that we will explore in the context of practical application within OASIS is the combination of modules of different complexities, so as to minimise the use of more expensive computational techniques without being forced to dispense with their capabilities altogether. Currently, when sophisticated first-order logic ontologies are to be used in practical applications, it is common practice to produce a simplified version in, for example, OWL-DL that may be used with a description logic reasoner: this is a time-consuming and expensive task, typically done by hand. Moreover, certain capabilities of the full ontology are necessarily lost. Heterogeneous specifications offer the following alternative solution.
The heterogeneous specification can use an efficient reasoner for the less complex parts of an ontology and only resort to a more complex reasoner if some concrete reasoning problem requires it. An example from the spatial domain would be an ontology that combines information about geographic administrative entities such as countries, towns, regions and streets with spatial information such as spatial extent: that is, one part of the ontology might specify a hierarchical representation of administrative units EU:France:Paris and another might specify geographical extent, the actual areas covered by those administrative units. Reasoning about these two types of entities is quite different: the hierarchical information can be expressed within a description logic and so falls within OWL-DL; reasoning about spatial extents and regions requires a different logic (e.g., the region connection calculus RCC: [Randell92], which is well known to lie beyond description logic). Combining these two specifications naively would force the entire specification beyond description logic, making it difficult to use for real services. Forcing the spatial information to fall within description logic by using a simplified version would mean in turn that the ontology is usable in practice but cannot reliably process information that relies on spatial extent. The OASIS heterogenous approach escapes this problem by factorising not only the ontology specification but automatically the kinds of reasoning tools that are used for each sub-ontology. There are efficient constraint-based problem solvers for regions and so a heterogeneous specification can maintain practical reasoning capabilities even for a complex ontology.
There are several more advantages to be explored for heterogenous ontologies in practical settings. For example, even though finding a relation between two complex ontologies might be computationally expensive, once that mapping is known, it may be possible to downscale the result so that it fits within a less computationally expensive framework. That is, the complexity is reduced to a compilation-time problem that is solved only once and then can be relied upon at run-time without requiring complex reasoning. Again, this allows us to build on the information maintained in sophisticated formal ontologies, in a way that is not available to simple ontologies while still maintaining a migration path to application.
OASIS will develop a new framework for heterogeneous ontology specification, building on this highly innovative perspective on ontological engineering. All of the ontologies developed within OASIS will be specified in this way and will make maximal use of modularity and diverse expressivity. Sub-ontologies will always be specified, so as to draw on more expressive resources as little as possible and only where the services they are to support need them. Techniques will be provided for re-using existing relationships between modules of any complexity so that already proved relationships do not need to be rediscovered. The application scenarios will provide explicit demands on the reasoning capabilities to be provided and will enable a concrete evaluation of the techniques developed. The entire framework therefore establishes for the first time a substantial real-world demonstrator both for the potential utility of axiomatised foundational ontologies and for concrete development paths for moving highly generic results from formal ontology into evaluable services. The heterogeneity of the framework also simultaneously guarantees that a basic level functionality will always be supported—for example, ontologies that fall within description logic in any case will provide a minimum of expressivity that services can rely on when transferring information. In addition to this, the OASIS scenarios will pick out and highlight enhancements in service capabilities, including adaptivity for the differing age groups, that are only supported when more complex modules can also be drawn on in the interoperability. In this way the project will illustrate and document precise classes of application cases where heterogeneity is the preferable approach.
The usefulness of the proposed ontological framework and the improved and new services that it enables will be demonstrated by means of the following applications:
- Domotic monitoring for elderly people: The application solution addresses the problem of proper integration of medical body sensors (from various vendors) and related health information with user preferences. This application proposes a system that can combine different levels of heterogeneous knowledge into a common hyper-knowledge in the form of a hyper ontology that can help elderly users not only to be monitored in their houses, but also custom actions to be taken, according to users special needs and preferences, in case of emergency.
- Autonomous mobility and smart workplaces for elderly people: The OASIS solution provides interconnected mobility services, targeted to the elderly user’s profile, as well as appropriately customised and secure access to their working environment.
From the developers’ point of view, plug-ins should be implemented that update the functionality of each module to be made compliant with the main hyper-ontology. This action does not require great changes (after all no vendor is willing to change its products from the ground up). The definition of the functionality of each module should be ontologically described and ontology should be added. The integration of ontologies will be feasible by aligning/relating the concepts of each sub-ontology with a main hyper-ontology that can combine smaller ontologies in a common manner within a common ontological framework. This will rely on the hyper-ontology’s support for flexible mappings between ontological components to minimise changes within any single contributing sub-ontology. Principles for best practice for constructing ontological descriptions, compatible with the hyper-ontology, will be developed and circulated during the project.
OASIS-defined agents will be able to leverage freely off the combined ontologies, avoiding explosions in complexity. A contribution of the COF will be to identify appropriate reasoners for diverse kinds of specifications: thus agents will be able to make use of reasoning components flexibly and as required for the particular problems they are dealing with. From the user’s point of view, the benefits of these capabilities are tangible. Without having to change services or habits, the user can combine various domains of knowledge in a unified manner and take advantage of various combinations of rules and events that weren’t feasible without a common language between various domains of information. Also, the user can combine already available equipment, which apparently could not be easily combined together.
The method for achieving the qualitative improvement in functionality for services for the elderly pursued in OASIS can be summarised as resting on three component technologies that have not been combined previously: (i) Web-based services using restricted ontologies for access, (ii) formal foundational ontologies for capturing the semantics of the devices and services to be inter-related more appropriately, and (iii) the methods and standards developed for structured software engineering in terms of algebraic specifications. This unique composition guarantees that OASIS can begin immediately to relate services as already pursued for hybrid-web services while successively augmenting their functionality, as more sophisticated inter-ontology relations are developed.
