|
Project Information
|
This is the source code for my PhD research. It is an active project. Please report any bugs to me: Luke.Steller@gmail.com See my research website: http://hercules.infotech.monash.edu.au/~lukes/ My PhD dissertation focused on leveraging Semantic Web technologies to perform semantic inference driven matching on-board a resource constrained mobile device. Semantic based matching provides greater accuracy than keyword / interface based approaches. Reasoners, which are used to perform semantic matching, are typically resource intensive. Therefore, most current mobile semantic matching approaches delegate processing to external / remote high performance servers. However, there is an emerging class of scenarios which can benefit from performing semantic service matching on-board the mobile device. Consider the example of a mobile user wishing to search for products / services using their mobile phone in a foreign city centre. Provision and maintenance of external servers which perform matching remotely would be more costly than performing matching on-board the user's own device. Additionally, continuous network access is not always available and tends to drain batteries more than CPU usage. There are also privacy concerns with transmitting sensitive data about a user's shopping preferences and habits to a remote third party server. In order to support on-board mobile matching I proposed mTableaux which enables mobile semantic reasoning by significantly improving efficiency without significantly reducing result accuracy. Additionally, current semantic reasoners operate on the premise that the matching process must be completed in full in order to obtain a result. To facilitate a more adaptive approach, I proposed an innovative capacity to provide priority based, incremental matching of a request to a semantic service description. This process can be interrupted depending on constraints such as user preferences, time and availability of resources, to provide a match result based on the matching completed. My strategies are based on the Tableaux semantic inference proof algorithm, which is widely used by current open source and commercial semantic web reasoners such as Pellet, FaCT++ and RacerPro. I implemented my proposed strategies as an extension to the Pellet reasoner. I conducted extensive experimental performance evaluations which clearly demonstrated that my strategies significantly improved response time and effectively prioritised the matching of conditions in a user request based on the importance of these to the user, to enable incremental matching. |