Applied Artificial Intelligence – Software

An important driver for any application of artificial intelligence is new and innovative code. Of particular interest are ideas for powerful algorithms (or heuristics) which might be broadly applicable to many applications. But we want to go a step further; to break out of the “customised program for a specific application” mindset and begin finding new ways to reuse code for new applications. We welcome applications for postgraduate study from software developers with the daring to break new ground in artificial intelligence.

Current Projects

The School’s Applied Artificial Intelligence Laboratory is the new location for our group, which is now seeking research students. This laboratory is being purpose-built by academic staff and students into a very dynamic, creative and fun workspace for new ideas. It is equipped with a broad range of up-to-date computers, software development tools to support our projects as well as a number of games and simulation machines for inspiration and entertainment applications.

Scripted Avatars in a Virtual Environment

A problem for many locales in virtual reality communities is the “ghost town” affect, in which the scarcity of human-driven avatars among the built environment creates a somewhat unsettling illusion of abandonment. This project aims to overcome this problem in the Second Life campus of Murdoch University by adding a population of “movie extras” – attractive human-like avatars driven by automatic programs that move and act to create a general ambience of metropolitan activity. The challenge here is to create realistic motion and social behaviours for these. To make the illusion really compelling – not to say useful – we are adding a simple question-answering capability to such “extras” so that they can – at present by linking them to an A.L.I.C.E. chatbot, running on a remote server. A.L.I.C.E. is a long-standing open source project which allows the user to write custom AIML scripts, making it easy to control the answers given by the “extras” to a range of English questions. A.L.I.C.E scripts that could welcome visitors, help “tour guides” take them on tours to different points of interest on the island and inform them of current or forthcoming events. The scripts drive the speech of the “extras” on the island, so that if a visitor stops one of them, he/she could have a useful, realistic and friendly conversation.

Daydreaming through a Database of Emotional States

Some aspects of human cognitive processing across experiences stored in episodic memory seem quite different from conventional artificial reasoning by logical rules, such as that seen in case-based reasoning systems. One difference is that in humans, linkages between particular memory episodes can apparently be made in a number of qualitatively different ways, ranging from associations to emotional connections. These form the familiar recollective chains of memories we call daydreams. Free-association from one memory to the next does not appear to be well-described by rules of the kind commonly used to describe or simulate logical inferences. Efforts to enable computers to deal with representations of narratives in new ways could benefit from sequential indexing of this kind, provided that the conceptual representations are rich enough, and that a way can be found of modeling the emotional impact each elicits. A conceptual-graph-based FGP (Fetch, Generalise, Project) machine using a knowledgebase of archetypical narratives enhanced with representations of emotions is now being used to discover how such emotive “memory-walks” can be computed without using rules.

Principle Researcher – Graham Mann

Applications of Memory-Based Learning to Spoken Language Interface Construction

This project aims to build a new and better tool for making Spoken Dialogue Systems – those which respond appropriately to a wide range of spoken input commands. The tool uses a novel Hierarchical Hidden Markov Model (HHMM) based algorithm to quickly learn question-and-answer pairs (or command-and-response pairs) from examples. The claim is that this architecture is a fast and intuitive way to construct a SDS capable of learning to answer questions and respond to commands to control connected equipment. As well as evaluating the built language-controlled applications (such as a simulated home automation system) as effective products for users, the tool is being tested for speed and ease of use by developers of such products.

Principle Researcher – Owen Lamont

Possible New Projects

Research students are encouraged to develop ideas of their own in consultation with academic supervisors within the Applied Artificial Intelligence group. Possible new projects include:

  • Autogeneration of affective states from attitudinal policies
  • Expansion of case based reasoning paradigm to an experience based paradigm
  • Automation of AR Drones to conduct simple missions