Commonwealth Bank gets machine learning to solve abusive messaging issues
Eighteen months after finding a large number of abusive messages attached to customers’ transactions with each other, the Commonwealth Bank of Australia (CBA) is implementing machine learning to detect abusive behaviour in transaction descriptions within the CommBank app and Netbank.
In June 2020, CBA had found more than 8,000 transactions received by customers to contain “potentially abusive” messages. “All genders were sending and receiving these messages, but the nature ranged from fairly innocuous ‘jokes’ using profanities to serious threats and clear references to domestic and family violence,” said Catherine Fitzpatrick, CBA’s general manager of community and customer vulnerability, at the time.
CBA started blocking transactions that contained offensive language in transaction descriptions. The bank found at the time that often transactions—for amounts as low as $1—would be used for the sole purpose of harassment.
Among 100,000 blocked transactions, what the bank called the ‘new AI model’ has identified 229 unique senders of potentially serious abuse, which were then manually reviewed to determine severity and the appropriate action required from the bank.
This AI model combines a range of machine learning techniques and adapting capability developed by Google. “For example, natural language processing (NLP), is used to transform all transaction descriptions sent between individuals, and the results are assessed for many factors, including how toxic or emotional the language is,” said CBA chief decision scientist Dan Jermyn. “This is then combined with further data points to understand patterns of transactional behaviour between customers. Billions of transactions can be assessed, from which we are able to predict those cases that are highly likely to require attention. The use of machine learning also allows us to adapt to results and become increasingly effective over time.”
Smart city tally ranks 5 Australian cities
Five Australian cities have been ranked in a smart city tally that considered 150 cities across the world across four areas. The Cities of the Future Index was performed by EasyPark Group
Cities were ranked based on digital life, mobility innovations, business tech infrastructure, and sustainability. They were also divided across three size categories of metropolitan areas: populations more than 3 million people, populations between 600,000 and 3 million people, and populations between 50,000 and 600,000 people.
Canberra (45th) and Wollongong (50th) made it to the top 50 in the category of populations between 50,000 and 600,000 people. Both Canberra and Wollongong had their highest scores across healthcare innovation, e-payments, green buildings, and waste management.
Adelaide ranked 44th in the top 50 of cities with populations between 600,000 and 3 million people. Its best score was in government adoption, which isn’t a surprise considering it is the Australian capital, but also did well in e-payments, green buildings, and waste management.
For those with population with more than 3 million people, both Melbourne and Sydney were among the top 50, coming in at 27th and 30th, respectively.
One of the points analysed was business innovation, which was one of the Australian cities’ lowest scores. That category evaluates startup activity in healthcare, internet services, financial services, lifestyle services, and media. The low score means that, compared to other cities, the Australian ones had less startup activity.
Australia and Finland to exchange supercomputer information
The Pawsey Supercomputing Research Centre and CSC–IT Center for Science in Finland have signed a memorandum of understanding to exchange of information regarding supercomputers to achieve strategic global objectives.
Both centres want to help promote research and development, solve highly advanced computing solutions, and enhance the productivity of global research and efficiencies. As part of the partnership, the centres will also focus on identifying opportunities to improve the operational efficiency of each party, and to provide enhanced support for globally distributed research teams.
Copyright © 2021 IDG Communications, Inc.