The use of ontologies in contextually aware environments

The use of ontologies in contextually aware environments Ian Millard, David De Roure, Nigel Shadbolt Intelligence, Agents, Multimedia Group, University of Southampton 1 st International Workshop on Advanced Context Modelling, Reasoning and Management

My background… Ph. D student at Southampton, supervised by Professors Dave De Roure and Nigel Shadbolt. Have interests in areas covered by two major IRC projects – AKT is concerned with all aspects of the knowledge lifecycle, from capture to reuse, with emphasis on semantic technologies and knowledge services Equator focuses on the augmentation of physical and digital environments, through a range of novel devices and interactions 1 st International Workshop on Advanced Context Modelling, Reasoning and Management

‘Ambient Intelligence’ Aim to create an environment which can pro-actively assist users going about their everyday tasks This will require some understanding of the people occupying the space, and the activities undertaken within it In the first instance, we are considering the presentation of relevant information to users at the right time and in the right place. This implies we need a level of contextual awareness – an understanding of where users are, and what they are doing 1 st International Workshop on Advanced Context Modelling, Reasoning and Management

Where? Determining the location of people and devices is non-trivial Likely that several different techniques and/or sensor technologies will be required Location ontology can model the layout of buildings and more abstract exterior spaces, symbolic emphasis over spatial. Ontological model offers useful inference, and permits data from different sources to be combined easily Dynamic RDF repository provides a flexible resource on which further applications and services may be based 1 st International Workshop on Advanced Context Modelling, Reasoning and Management

Examples Gains through basic ontological properties eg 1, inverse properties loc: zepler. L 1, loc: is-below, loc: zepler. L 2, loc: is-above, loc: zepler. L 1 eg 2, transitivity of is-located-in ecs: icm, loc: is-located-in, loc: zepler. lec 1 ecs: icm, ecs: icm, loc: is-located-in, loc: is-located-in, loc: zeplerbldg loc: maincampus loc: univofsoton loc: hampshire loc: england 1 st International Workshop on Advanced Context Modelling, Reasoning and Management

Examples Custom rules allow more complex inference rule 1: (D, loc: is-located-in, L) && (P, dev: is-using-device, D) =>(P, loc: is-located-in, L) dev: pda, dev: has-user, ecs: icm, dev: is-using-device, dev: laptop dev: pda, loc: is-located-in, ecs: zepler. lec 1 ecs: icm, ecs: icm, loc: is-located-in, loc: is-located-in, loc: zeplerbldg loc: maincampus loc: univofsoton loc: hampshire loc: england 1 st International Workshop on Advanced Context Modelling, Reasoning and Management

Owl. Srv To provide the storage, inference and query capabilities required for development, 'Owl. Srv' has been created Based on Jena libraries, Owl. Srv offers persistent storage of RDF data and OWL ontologies, and permits dynamic updates Inference is performed over the data stored, in accordance with the loaded ontologies and a set of custom rules Query interface based on RDQL over HTTP Experimental system – limited performance and scalability 1 st International Workshop on Advanced Context Modelling, Reasoning and Management

What? While location is an important factor in contextual awareness, but we must also consider what people are doing Creating a single ontology to represent some notion of generic context is likely to be difficult and of little use Most working days can be conceptualised as a series of events or tasks, hence we are looking at task modelling Using templates we can hierarchically categorise common tasks and events, and propagate value inheritance through custom inference rules 1 st International Workshop on Advanced Context Modelling, Reasoning and Management

Examples Value inheritance through custom rule 'is-instance-of' foo: Lecture, task: is-instance-of, task: Generic-Event foo: CM 143 -Lec, task: is-instance-of, task: Lecture task: requires-resource, foo: CM 143 -Syllabus task: has-interest, foo: Computer-Networks task: is-presented-by, ecs: Dr. Smith foo: CM 143 -Lec-1, foo: CM 143 -Lec-1, task: is-instance-of, foo: CM 143 -Lec task: requires-resource, foo: CM 143 -Notes-1 task: has-next-instance, foo: CM 143 -Lec-2 task: requires-resource, foo: CM 143 -Syllabus task: has-interest, foo: Computer-Networks task: is-presented-by, ecs: Dr. Smith foo: CM 143 -Lec-2, foo: CM 143 -Lec-2, task: is-instance-of, foo: CM 143 -Lec task: requires-resource, foo: CM 143 -Notes-2 task: requires-resource, foo: CM 143 -Syllabus task: has-interest, foo: Computer-Networks task: is-presented-by, ecs: Dr. Smith task: has-prev-instance, foo: CM 143 -Lec-1 1 st International Workshop on Advanced Context Modelling, Reasoning and Management

Example Application Experimental work looking at the feasibility of some of these ideas Designed to assist a senior academic, by presenting a task oriented view of the day Several ontologies used along with custom inference rules to identify events and papers of interest to the current task Semi-automated collation of information shows potential to be useful, though currently uses limited hand crafted data 1 st International Workshop on Advanced Context Modelling, Reasoning and Management

Future Work Continue with current activities, expanding from single client application into pervasive environment Main focus 'Contextual Message Delivery' – sending notifications to users in the most appropriate manner Will require further modelling of importance, ‘interruptability’ Interpersonal awareness – where, what and with who Integration/use of external services Various methods for automated data population 1 st International Workshop on Advanced Context Modelling, Reasoning and Management