Photo credit attributed to the youtube video of Christina Aguilera's performance of the song It’s A Man’s Man’s Man’s World 2007 Grammy Awards.

Member-only story

Topical affinity across Family Violence policies and communication platforms

Arezou Soltani
7 min readJul 14, 2021

“Respect, dignity, choice, equality of opportunity and justice are fundamental to the safety, economic security and status of women in Australia.” This is the opening line of the 2021–22 Women’s Budget Statement indicating the Australian government’s commitment to reduce/prevent all forms of family and sexual violence as steadfast. In this post, I recap the SOPHIA project in which, we harnessed a suit of Natural Language Processing (NLP) techniques to assess the extent to which changes in public attitudes and engagement could be detected in response to Family Violence (FV) policies. Please note the content of this pitch is solely the responsibility of the author and does not necessarily represent the view of the State of Victoria.

From crawl to walk

Since our initial meetings with Victoria Department of Premier & Cabinet (DPC), we realised, for this project, acquiring subject matter expertise was as much critical as the data science skillset. From my own experience, many questions of societal importance are causal questions and often govermental agencies are interested in predicting the effects of their actions, rather than understaning how things are related in a world without any intervention. In our context, the major question we faced was the effects of government polices on addressing FV. For these type of problems, the data-centric thinking that is prevalent in machine learning research, according to which the secret to rational decisions lies in the data alone, is not sufficient [1].

Instead of rushing into analytical part, we made sure subject matter experts are on board with us: A steering group of FV experts, policymakers and social workers, convened by DPC, guided our understanding of FV discourse from the beginning throughout the analysis, and gave us feedback on findings. An immediate data source recognised from these meetings was the Royal Commission into FV. Between February 2015 and April 2016, the Commission held 44 group consultation sessions with 850 people, received over 1,000 written submissions, and heard from 220 witnesses over 25 days of public hearings, made 227 recommendations in total.

--

--

Arezou Soltani
Arezou Soltani

Written by Arezou Soltani

Principle Machine Learning Engineer at Atlassian | Recipient of VentureBeat’s Women in AI Leadership Award 2021 | I write about some of my interesting projects

No responses yet

Write a response