The use of machine learning to analyze job advertisements for doctoral employability

Abstract

The nature and extent of the demand for research capable workers, is a topic of intense concern locally and internationally. With around 60% of graduates in Australia finding employment outside of academia on graduation, PhD programs are under increasing pressure to be relevant to the contemporary workplace beyond the walls of the academy. However, as yet, there is very little research on exactly what industry needs are as of the discussion with industry results in recommendations based on anecdote rather than data. This study aims to fill this gap by analysing a large data set of job ads to see what employers outside academia really want from graduates.

This research builds on an exploratory study which analysed job adverts for roles specifying a PhD as a required or desired criteria in academic roles (Pitt and Mewburn, 2014). By focussing on what is actually stipulated as required for these roles at the time of advertising them, an alternative picture emerges of what employers really want in PhD-qualified employees. This next stage will use machine learning to investigate (and, should this project be successful, track) the demand for advanced research skills amongst Australian industry sectors. To do this we plan to systematically explore, annotate and catalogue job advertisements (drawn at first from Australia's largest online employment marketplace seek.com.au) advertising for highly paid knowledge workers. Data drawn from the SEEK database will be processed using language analysis and classification algorithms, allowing the development of a tool able to assess which Australian industry sectors are looking to hire researchers, and the skills they are looking for.

This paper reports on the first phase of this project which involved building and testing a robust ontology for describing PhD graduate skills and its application to classify textual job ads retrieved from SEEK. Classifications of three content experts are compared. By making this data visible to research students, research managers, businesses and government we can find ways to better connect Australian businesses with Australia's highly skilled research workforce and provide new directions for those engaged in supporting PhD students in their career post PhD. Show more