Ideas for final year or MSc dissertation projects
- Most of these projects are going to be related to Artificial Intelligence in some way, but I try to keep an open mind about what sort of technique will be most appropriate..
- All of these projects are going to need people who are comfortable writing code. In some cases I can provide links to existing tools, but these will probably need adapting for the particular research area.
New this year
- Using kinetic to control interactive evolutionary art
- Building chat bots for ... (Loads of stuff needs doing around the faculty or for my module)
- Depending on your coding , interactive evolutionary of "talking head" avatars for chatbots
- Evaluation of scratch/app builder for creating android games
- Automatic creation of software class models using metaphors from from how ants cluster eggs in a nest t (we have data and could use the netlogo multi-agent framework or jade or whatever).
- We have some really nice interactive AI-based software for creating and customising Image processing routines or web-page design.
There would be several nice possibilities to work on this, or on things related to it.
For example you might look at:
Building intelligent adaptive avatars for e-learning situations.
- An interactive ai based applet for CSS design
- Methods for augmenting the "intentional" stream of user inputs with meta-data about their activities to let us know about things like are they tired?, how engaged are they, or frustrated?
Types of meta-data we might consider would be response times, or descriptors of their mouse movements, or using web-cams to record and classify their body posture.
I use on-line tests for assessment, and to provide people with self-assessment tools quite a lot, but at the moment they’re dumb in the sense that they just present a fixed list of questions.
- We've got a number of interested groups who want to use intelligent "bots" to help train people in interviewing etc.
- The idea is to provide "bots" which can respond to students showing emotional traits in their movements (nervousness, confidence etc)
- and /or intelligently interpreting questions.
- We did some good exploratory work using Linden scripting in second life a couple of years ago , but there is lots of scope for improvement and building a solid s/w platform
It would be really nice to look at how we could apply artificial intelligence techniques to decide what questions to present a learner with next. For example:
Evolutionary algorithms are a way of generating good solutions to problems, based on the idea of Darwinian evolution.
- Maybe the system could be more likely to ask more questions on topics where the learner hasn’t done well?
- and correspondingly less likely to ask questions on topics they clearly understand.
- Equally, there are various ways of asking a question e.g. true/false, multiple choice, matching, multiple answer, filling in blanks in a sentence.
Different questions suit different people, could we make a system that adapted these on-line to a users preferences in order to improve their learning experience.
- I can provide the students to test the results, and the questions, butthe student would need
- ability to work with data base of questions ( I can provide these)
- ability to analyse problem and work out what kind of AI would be applicable here
- ability to write code to run an adaptive system modelling a user’s knowledge and experience
- ability to link these together and maybe present via a GUI
- In an ideal world I’d like to link this all in with Blackboard, but that might be too much
Increasingly people are realising that it is a good idea to add some elements of individual learning to evolution to form so-called “memetic algorithms”.
In nature most higher species learn/adapt individually as a result of trial and error, but also learn from their parents, peers and teachers – so called “social learning”.
There are a number of possible projects in this area such as:
To be continued
- Building an on-line system for detecting road traffic signs. Starting off with single images then moving on to videos.
Quite a lot of useful software tools out there that could be integrated.
By the end of this you would have good image processing skills and some knowledge of AI techniques for object classification.