By Carrie Cox
For all its reliance on systems and technology, artificial intelligence would simply collapse without the scaffolding of human knowledge: what we know informs what we don’t know and what we might ultimately benefit by knowing.
Given that the vast majority of human knowledge is captured in text — our reports, diaries, observations, conversations and literature — the development of what is known as ‘natural language processing’ has been critical to the early evolution of AI. Without it, the value of AI and the big data it feeds upon would be significantly diminished. There’d be lots of ‘artificial’, not so much ‘intelligence’.
Natural language processing — essentially, the AI field that centres on the interaction between computers and humans through language — is a field of learning that has become a specialty for Wei Liu, an Associate Professor within UWA’s School of Physics, Maths and Computing. She heads up the University’s Natural and Technical Language Processing Group, an award-winning research team that collaborates with other computer scientists around the world, while also delivering real-world data solutions for WA industry.
Recent projects have included work with the Geological Survey of Western Australia to produce ‘knowledge graphs’ based on geological publications to better understand the mineralisation processes of critical minerals and ultimately improve exploration prospecting.
Another project, in partnership with a Perth-based environmental consultancy, combined large datasets to streamline green assets management and evaluation for mining rehabilitation and environmental applications to city councils. And another used data supplied by the Department of Mines, Industry Regulation and Safety to deep-analyse safety logs and better understand how workplace injuries could be prevented.
“AI can do these sorts of analysis so much faster and more accurately than human eyes,” Associate Professor Liu says. “We’re helping organisations make data-driven decisions and data can often deliver you simple insights that humans overlook.
“What’s particularly exciting is that this information not only helps people better understand what is happening within their organisation but also what might happen down the track. In the case of safety, for example, that means using data-driven insights to prevent future injury.”
Much of Associate Professor’s Liu’s expertise addresses the challenge of ‘interoperability’ within big data analysis. This is essentially the ability of different data sources — for example, social media logs, sensors, databases and various types of software — to speak the same digital language. Without interoperability, sense can’t truly be made and any data analysis will be inherently flawed.
“This remains a challenging problem in data analysis,” Associate Professor Liu explains. “Knowledge is always captured with specificity and the choice of technical language facilitates communication.
The other day my husband and I were watching ducks burying their heads into the water with their rear-ends up and my husband said ‘There must be a word for that’. I described the action for ChatGPT and it immediately told me the duck’s action was called a ‘dabble’. Isn’t that terrific?
Associate Professor Wei Liu, School of Physics, Maths and Computing
“My early research ventured into a bottom-up approach to the development of ontologies (explicit representations of knowledge and language) where I would talk to individual companies to understand their specific naming system and then use the equivalent of a crawler to extract commonalities and build up a more universal language that enables collaborative sharing
“Numerical data is easy to capture but text-based AI is much harder – how do you predict things from sentences? Yet human knowledge is about 80-90 per cent captured in text, so the real-world applications of natural language processing are many. Basically, when you’re talking about any knowledge-intensive industry or economy, you’re talking about text.”
Breakthroughs in ‘deep learning’ (computerised learning inspired by the workings of the human brain) just over a decade ago have super-charged Associate Professor’s Liu’s research and produced everyday AI applications like ChatGPT. But knowing how the sausage is made hasn’t diminished her own delight in the AI wonders now at our disposal.
“I find it all very exciting,” Associate Professor Liu says. “The other day my husband and I were watching ducks burying their heads into the water with their rear-ends up and my husband said ‘There must be a word for that’. I described the action for ChatGPT and it immediately told me the duck’s action was called a ‘dabble’. Isn’t that terrific?
“I know many people have concerns about AI and there are some big philosophical questions around it, but I am optimistic that it will ultimately lead to positive growth — it is already saving lives — because civilisation wants to progress and because we are in charge. We are still the ones driving.”
Read the full issue of the Winter 2024 edition of Uniview [Accessible PDF 12MB]