School of Computer Science / Centre for Language & Communication Studies
Supervisors: Dr Carl Vogel / Prof. Nick Campbell
Making Computers More Human-like
The state of the art of computer speech is not exactly the best. It’s very flexible, but not particularly natural sounding. This unnatural nature makes it harder to understand. Listening to computer generated speech is in some ways a lot more like reading than listening, in that it’s a skill and it takes a lot of effort. For most people, that’s not that much of a problem. But once something like Alzheimer’s starts taking your abilities, it gets harder.
My research is about making computers sound more human. That way, instead of having to learn a new skill, you can rely on all the ones you learned as a child. The human mind notices dozens of cues which it then uses to understand what people are saying. The more of those cues that we can identify and insert into speech synthesis, the easier it should be to understand.
In the future, we hope that personal computers will be more like personal assistants. But, unless we have a better speech synthesis programme, it’s still going to feel like work. I think we would all prefer if the current speech programme is replaced by a better generation of computer voices – because who really wants to listen to a voice that annoying?!