Natural speech Automated Utility for Mental health
Lead Research Organisation:
University of Cambridge
Department Name: Engineering
Abstract
Promotion of mental well-being is at the core of the World Health Organisation's action plan on mental health 2013--2020, with particular emphasis on the prevention of mental illnesses.
Indeed, prevention has long been neglected: if we were to make an analogy with dentistry, the state in mental health is such that we know how to treat caries, but we have yet to discover toothpaste.
Although there is research to suggest that internet-based therapy can be beneficial, there has been little progress on automated mental health advice systems. In the last decade, machine learning has made a huge impact on various areas including spoken dialogue systems.
Still, the application of statistical spoken dialogue systems has so far been limited to simple information-seeking tasks. Here we propose NAUM---Natural speech Automated Utility for Mental health---, a purely data-driven spoken dialogue system that can be used for maintaining mental well-being. Mental health experts will work on developing NAUM's knowledge, its behaviour will be optimised by novel reinforcement learning algorithms and it will support spoken interaction.
This ground breaking research will bring the potential of machine learning in spoken dialogue modelling to an application which has a clear benefit for society. NAUM will provide anonymous support that can be accessed by anyone, any time, anywhere, for free.
Indeed, prevention has long been neglected: if we were to make an analogy with dentistry, the state in mental health is such that we know how to treat caries, but we have yet to discover toothpaste.
Although there is research to suggest that internet-based therapy can be beneficial, there has been little progress on automated mental health advice systems. In the last decade, machine learning has made a huge impact on various areas including spoken dialogue systems.
Still, the application of statistical spoken dialogue systems has so far been limited to simple information-seeking tasks. Here we propose NAUM---Natural speech Automated Utility for Mental health---, a purely data-driven spoken dialogue system that can be used for maintaining mental well-being. Mental health experts will work on developing NAUM's knowledge, its behaviour will be optimised by novel reinforcement learning algorithms and it will support spoken interaction.
This ground breaking research will bring the potential of machine learning in spoken dialogue modelling to an application which has a clear benefit for society. NAUM will provide anonymous support that can be accessed by anyone, any time, anywhere, for free.
Planned Impact
The main objective of the project NAUM is the development of a spoken dialogue system for supporting people at risk of developing depression. To achieve this we set ourselves the following three milestones:
- Development of tools for a mental health ontology that can be used inside a spoken dialogue system
- Development of novel reinforcement learning algorithms that can model complex interaction
- Data collection of human-computer mental health therapy data
Successful development of such a system opens new avenues that go beyond this project. Such a system could then be used in clinical trials aimed at preventing the development of mental illness. Other research questions could be investigated, such as can the ontology be automatically learned from CBT manuals and can the dialogue system be multi-modal and support affective interaction.
- Development of tools for a mental health ontology that can be used inside a spoken dialogue system
- Development of novel reinforcement learning algorithms that can model complex interaction
- Data collection of human-computer mental health therapy data
Successful development of such a system opens new avenues that go beyond this project. Such a system could then be used in clinical trials aimed at preventing the development of mental illness. Other research questions could be investigated, such as can the ontology be automatically learned from CBT manuals and can the dialogue system be multi-modal and support affective interaction.
Organisations
People |
ORCID iD |
| Milica Gasic (Principal Investigator) | |
| Mike Crawford (Co-Investigator) |
