Natural Language Generation for Low-resource Domains

Project Description

In the UK, around one in four people suffer from a mental health problem every year [1]. Mental well-being can affect all aspects of our lives, including our work, relationships, and family lives. A lack of personalised information along with the mental health stigma creates barriers to support. Failing to deal with our mental health can have a severe impact on expectancy and quality of life. To combat that, promotion of mental well-being, management and prevention of mental health illnesses has been identified as a core priority of the World Health Organisation's mental health action plan 2013-2020 [2] as well as in NHS's "Five Years Forward View" for mental health [3].

Younger people are particularly less likely to seek help when facing mental health challenges [4], therefore, it is of vital importance to create a safe space for younger people that empowers them to seek advice and information regarding mental well-being. 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, artificial intelligence has made an impact on the creation of novel natural language interfaces such as personal assistants. Personal assistants can offer a way to access the young population in a way that is familiar to them, through written texts and by being able to access them anytime. Although recent advances in understanding natural language have made it possible to accurately predict the meaning of users' utterances and hence accurately inform the personal assistants' actions, responding to it in natural language remains a bottleneck for dialogue systems/personal assistants.

Current response generation techniques are heavily based on pre-specified templates that limit language coverage. Generating fluent responses is heavily dependant on example dialogues, which are not available for this domain. To address this challenge, the project will firstly develop natural language generation techniques that are able to learn from limited resources. Secondly, in order to increase young people's engagement with the personal assistant, the generated texts should remain interesting and novel even after several dialogues. Therefore, the second goal of the project will be to develop approaches that can generate text that is variable, with the aim to enhance people's experience and increase engagement. Finally, the project will aim to adapt the generated text to the users' emotional state, since emotional awareness and empathy can contribute to building trust between the personal assistant and the young people.