Catalyst: Strategic – New Zealand-Singapore Data Science Research Programme

Joining the researchers and science leaders in both data science and neuroscience at the University of Auckland, New Zealand, and Nanyang Technological University, Singapore, this project aims to understand and locate deep connectivity patterns in the human brain with cutting-edge graph analytics tools.

Understanding the human brain has both economical and social importance. It helps the identification of subtle differences (discriminative patterns) between data from brains with early symptoms of neurodegeneration (Alzheimer’s and Parkinson’s), mild traumatic brain injury (mTBI), etc. A report from Government Inquiry into Mental Health and Addiction of New Zealand in 2019 shows one in five people experiences mental health and addiction challenges at any given time while the annual cost of serious mental illness in 2019 is 12 billion which takes 5% of the gross domestic product of New Zealand. Similar scenarios also apply to other countries around the world. mTBI is also a significant issue both in New Zealand (NZ) and worldwide. There are 749 cases per 100,000 person-years of concussion in NZ, greater than any other country in the developed world. ACC statistics show that nearly 14,000 people were treated for TBI, at a cost of $83.5 million in 2015; Māori recorded the highest total and mean cost per ACC claim compared with any other ethnic group. Annual NZ figures from ACC (data request #46858) also show a 32% increase from 2010 to 2017 in total new concussion-related claims. This demonstrates the necessity for a deeper understanding of the human brain to allow early diagnosis and interventions of subtle TBI induced changes in the brain network.

In the project, Magnetic Resonance Imaging data that reflects both structural and functional connectivities of the brain will be collected from both public sources and local teams of Singapore and New Zealand. The data will be processed and nalysed with novel multiple parcellations discovery and group-based graph analytics for brain networks. The outcome of this research will benefit the health of the individual and the society, including Māori, by early diagnosis and intervention of mental illness, neurological and aging-related disorders/diseases, etc. It will also contribute to the establishment of a healthy aging society in both New Zealand and Singapore, as well as in the global world, with quality elderly life and reduced healthcare costs.

Project team contacts

• Dr Miao Qiao (Science Leader) | The University of Auckland, New Zealand
• Assistant Professor Yiping Ke (Principal Investigator) | Nanyang Technological University, Singapore
• Professor Balázs Zoltán Gulyás (Co-Investigator) | Nanyang Technological University, Singapore
• Dr Yun Sing Koh (Key Researcher) | The University of Auckland, New Zealand
• Associate Professor Alan Defeng Wang (Key Researcher) | The University of Auckland, New Zealand
• Dr Samantha Holdsworth (Key Researcher) | The University of Auckland, New Zealand
• Dr Vickie Shim (Key Individual) | The University of Auckland, New Zealand
• Associate Professor Miriam Scadeng (Key Individual) | The University of Auckland, New Zealand

Te reo Māori, Malay, and Singlish, all play central roles in the culture and identity of the Māori and Singaporean communities. Both countries are committed to revitalising and promoting these languages, providing access to the vast indigenous and local knowledge that they express.

Recent advances in data science and deep learning for natural language processing (NLP) have opened up exciting new possibilities in major languages such as English and Chinese. However, there is no software system that can systematically integrate listening to, speaking, and reading in less widely used languages such as te reo Māori, Malay, or Singlish.

This project will develop novel NLP techniques in machine translation, Q&A, and novel speech processing techniques required to create an intelligent conversational Q&A system in te reo Māori, Malay, and Singlish. It will explore ways in which AI systems can gain knowledge expressed in these languages, a process called “knowledge capture”, broadening the range of cultural input to this increasingly important technology. The Q&A system integrates listening to, speaking, and reading te reo Māori/Malay/Singlish to benefit learners and users.

This project will be led by Prof Ruili Wang and be conducted by a collaboration between researchers at Massey University, the University of Auckland, the University of Waikato, and the Institute for Infocomm Research at the National University of Singapore (NUS). This jointly developed intelligent conversational Q&A system will provide state-of-the-art speech recognition, machine translation, Q&A, text-to-speech synthesis for language learning and revitalisation efforts, and more effective public systems and services. The application then has the potential to extend to other indigenous and vernacular languages.

The project strongly aligns with the national research priorities in both NZ (National Statement of Science Investment, Vision Mātauranga Policy, Maihi Māori 2017-2040 Strategy) and Singapore (National Artificial Intelligence Strategy). It centres on new knowledge creation in NLP and speech processing.

Project team contacts

• Professor Ruili Wang (Principal Investigator/Science Leader) | Massey University, New Zealand
• Professor See-Kiong Ng (Principal Investigator) | National University of Singapore, Singapore
• Associate Professor Stéphane Bressan (Principal Investigator) | National University of Singapore, Singapore
• Professor Michael Witbrock (Key Researcher) | The University of Auckland, New Zealand
• Associate Professor Te Taka Keegan (Key Researcher) | University of Waikato, New Zealand
• Professor  Huia Jahnke (Key Researcher) | Massey University, New Zealand
• Dr Darryn Joseph (Key Researcher) | Massey University, New Zealand

Singapore, the ‘City in a Garden’, embodies the ‘green city’ concept with over 7 million urban trees covering 700 km2. New Zealand, with 24% of its 270,000 km2 land covered in forest, also actively supports and promotes urban re-greening in many of its cities. Sustaining and enhancing biodiversity and healthy living environments are priorities for Singapore and New Zealand that require careful management of trees in urban areas and forests. Reliable information, models, and analysis of trees and their interaction with the surrounding environment are essential to inform management decisions. However, these are currently limited by the quality of available data, tools, and techniques.  

Leveraging our joint expertise in data science, remote sensing, and 3D modelling, we propose a proof-of-concept integrated methodology.  We will develop novel data-science methods for extracting tree species information from petabytes of multiresolution remote- sensing data to model tree species and their interactions with the environment, and subsequently analyse their socio-economic impacts. This work will form the basis for future research collaborations to enable further modelling, simulation, and analysis. In the long term, our work will empower and inform decision-makers on trees and environmental considerations for the greater benefit of both New Zealand and Singapore. 

Project team contacts

• Dr Jan Schindler (Science Leader) | Manaaki Whenua – Landcare Research, New Zealand)
• Dr Like Gobeawan (Principal Investigator) | Institute of High Performance Computing, A*STAR, Singapore
• Dr Alan Tan (Key Researcher) | Scion, New Zealand
• Professor Richard Green (Key Researcher) | University of Canterbury, New Zealand 
• Associate Professor Lee Bu Sung (Key Researcher) | Nanyang Technological University, Singapore 
• Professor Mengjie Zhang (Key Individual) | Victoria University of Wellington, New Zealand

New Zealand has one of the highest prevalence rates of depression worldwide. It accounts for half of the annual suicides and attempted suicides, particularly among 13-25 years old.  There is a need to develop methods for accurate diagnosis/prognosis of mental illness and suggest optimal interventions.

The main outcomes:

• Development of new machine-learning/AI methods for multimodal data modelling. 
• Better clinical intervention via early prognosis and diagnosis of mental health issues in at-risk youth.
• Developing personalised modelling for a better understanding of individual factors that trigger mental illnesses.

We aim at enhancing the accuracy of early detection/prediction of mental illnesses using a combination of different datasets including cognitive and genetic datasets.  We develop new methods for data fusion (a process of integrating multiple data sources) that provide machine learning with more consistent, accurate, and useful information.

Novel personalised modelling technique will be proposed, based on a new clustering approach for selecting a subset of informative datasets. This leads to create personalised profiling and enhance the classification and prediction of an individual cognitive state.

The integrated datasets will be modelled using a 3-dimensional structure of artificial spiking neurons to map the spatial information of the data while learning from the temporal patterns “hidden” in the longitudinal measurements. A combination of advanced techniques for multimodal data collection, data processing, data integration, and computational modelling will be proposed. These lead to solving the issue of data integration when dealing with several big dimensional data spaces (genes, biomedical and cognitive) from a cohort study. It also allows our proposed advanced deep learning techniques to learn from a more informative combination of features across multimodal data domains. This will potentially result in improved accuracy of outcomes. Besides the model accuracy, we will focus on improving the model interpretability. This refers to understanding the relationships between the model features and the predicted outputs, which has not been investigated in depth. The higher the interpretability of a model, the easier it is for someone to comprehend why certain decisions (output predictions) were made. This allows for knowledge discovery in the models and contributes to the understanding of interactions in the model that have controlled an output to occur. Findings would be theoretically relevant to a better understanding of mental illnesses in New Zealand and worldwide and can be used in the future to enhance service provision in New Zealand and Singapore.

Project team contacts

• Professor Nikola Kasabov (Science Leader) | Auckland University of Technology, New Zealand
• Dr Wilson Goh (Principal Investigator) | Nanyang Technological University, Singapore
• Dr Maryam Doborjeh (Key Researcher and Co-Investigator) | Auckland University of Technology, New Zealand