Spotlight on Women in Tech – Dr Cecile Paris
We’re talking to Dr Cecile Paris, Group Leader of the Knowledge Discovery and Management group at CSIRO’s Data61, about her most impactful projects, career advice for anyone looking to break into the tech industry, and using natural language processing to study the emotional health of populations in real time.
What led you to choose a career in tech? Tell us about your career journey so far.
I loved mathematics in high school, and one of my teachers introduced me to the programmable HP calculators. I thought they were so cool! Between maths and the logic in these calculators, I started studying computer science. After doing my undergraduate degree at UC Berkeley (California), my Honours Professor recommended me (without my knowledge) to the chair of the newly formed Computer Science Department at Columbia University (New York), who called me to offer me a scholarship to do a PhD. So, rather unexpectedly, off I went to New York. My first semester there, I took classes in Artificial Intelligence (AI) and in Natural Language Processing (NLP, a subfield of AI), and found both fascinating, so I switched to AI and NLP.
After graduation, I worked at the Information Sciences Institute at the University of Southern California (USC/ISI, California, USA), a prestigious research institute for AI research. At USC/ISI, my work focused on discourse planning for dialogue systems in the context of enabling knowledge-based systems to explain their behaviour. This became a major innovation in the computational embodiment of discourse coherence theories and User Modelling.
Late in 1993, I joined the Information Technology Research Institute (ITRI) in the UK, where I conducted research on the automatic production of software documentation and instructions (sometimes in several languages), building on my work in discourse planning. The automatic production of coherent multilingual documents (documents produced simultaneously in several languages from the same underlying representation) is a more accurate alternative to machine translation.
I was recruited to CSIRO in 1997 to start a research team in Natural Language Processing. I furthered research in discourse and personalisation and applying it to client projects. For example, we demonstrated the feasibility and appropriateness of tailored delivery in the government domain, a domain which is conservative by nature and in which personalisation had not been applied. The work paved the way for the Australian government agency responsible for the delivery of social and health services and payments, Centrelink, to implement “Payment Finder”, an application that, given some information about the user, outputs the relevant services and payments.
I led the Language and Social Computing team until I became a Group Leader. I was Data61’s Chief Scientist for 3 years, ending last September, and am an Honorary Professor at Macquarie University and hold several Adjunct Professorships. I am a Fellow of the Australian Academy of Technology and Engineering (ATSE) and of the Royal Society of NSW. I have supervised numerous PhD students and postdocs, some of whom won prizes, and I take great pride in seeing them flourish as researchers.
How did you end up at CSIRO’s Data61? What inspired you to join the organisation?
I was recruited to CSIRO to start a research team in Natural Language Processing. I was attracted to CSIRO, as Australia’s national science agency, as I understood that research at CSIRO would lead to real world impact, and that was very appealing. The Business Unit at the time was the Information Technology division, which, shortly after I arrived in Australia, was merged with the Mathematical and Statistics division to form the Mathematical and Information Sciences division. I then was part of the ICT Centre, and later Data61.
What are some of the projects you’re working on at CSIRO’s Data61? Can you tell us about some of the most impactful?
My most impactful Innovation has been the “We Feel” initiative, applying Natural Language Processing (NLP) to study the emotional health of populations in real time. Social media enables people to talk about themselves to share how they feel, to avoid feeling lonely, and potentially to get support, especially during hardship or in situations like bushfires or COVID-19. Yet the extreme volume and speed of messages, and the diversity of how people express themselves, seemed insurmountable obstacles for researchers to detect emotional health at scale.
Combining my expertise in AI, data mining and NLP allowed me to attack this problem head on, after initiating a partnership with The Black Dog Institute in 2014. We Feel is an NLP system that constantly monitors Twitter for (English) posts that express an emotion. We Feel is the first of its kind to monitor people’s emotions on Twitter in real time and at scale, which was a challenge given the amount of data (up to 45,000 tweets per minute). The system processes this flow of tweets and presents an analysis of the different emotions occurring around the world, using an “emotion wheel”.
The results are presented in an interactive dashboard where one can select regions of the world and see the emotions conveyed in tweets from that area. The system also provides an Application Programming Interface (API) that enables people to download the analytics data to be used for their own purposes (e.g. correlating moods with other data, identifying the population moods during specific periods, visualisations and art works etc).
We Feel, and our other work for the Black Dog Institute on detecting suicide ideation on Twitter, was the catalyst, and the enabling technology, for the Black Dog Institute to employ social media as a key source of information for its mental health research.
Six years on, We Feel has attracted almost half a million views, and the results of its analytics is being used, free of charge, by over 91,500 researchers, some of whom, in turn, contribute to improve mental health across the globe or research other societal phenomena. For example, the data has enabled novel work in statistical modelling at CSIRO’s Data61 and work by social scientists outside CSIRO to investigate the association between temperature and aggression in the context of domestic violence, while the software modules underlying We Feel have been embedded in several other software applications, and we have recently been approached by a psychologist in Victoria wanting to include some of the data from We Feel in an app aimed to support people with mental health problems.
What do you love about working in tech?
I love problem solving and looking for solutions to very hard scientific problems. I am motivated to use AI to solve real problems that impact many people’s lives. I enjoy working with scientists with different perspectives to obtain better solutions.
Why is gender diversity important in tech?
Gender diversity is important in everything! There has been a lot of research on the benefits of diversity (of all kinds) on productivity, teamwork, creativity, and outcomes in general.
It also seems appropriate that, given that women form ½ of the population, they should be ½ of the STEM workforce.
In your opinion, what’s the single biggest change that needs to happen in order to encourage more women to pursue careers in tech?
Breaking down social stereotypes – very early in children’s lives.
How can colleagues, organisations and industries within STEM better support and enable women?
Everyone should be actively aware of potential biases against women and actively seek to encourage and promote women. More women should be in these organisations and industries, as scientists (or STEM professionals), to provide more support and more role models.
What advice would you give to women and girls wanting to pursue a career in tech?
If you enjoy science, do it. It won’t necessarily be an easy journey, but it’s important we do it.