What can PostAc tell you about the job market?
PostAc was built using machine learning and natural language processing technologies. We first developed a ‘research skills annotation schema’: a series of statements that described the ideal PhD graduate. These statements were then applied to a large corpus of job ad texts. The human coders applied the research annotation schema to each of the job ad texts, highlighting text that was relevant in order to teach the machine what to look for. The human coders than made a judgement about the advertisement as a whole; categorising it as either high knowledge intensity, medium knowledge intensity, or low knowledge intensity (with high knowledge intensity signifying ‘PhD shaped’ jobs).
We then build a pilot demonstration system to produce histograms of the job ad rankings. In the diagram below, the x-axis represents increasing research skills intensity; the y-axis represents the number of jobs at each level. The further along the x-axis that a job falls, the more likely it is to require higher levels of research skills. The diagram below, shows a representation of all the jobs in the Seek.com.au data set as a histogram. The coloured section under the curve is a graphical representation of how many jobs we deemed ‘PhD shaped’:
As expected, some industries showed distinct patterns of higher and lower demand for research skills. For example, “trades and services” (Figure 2, below) has an abrupt drop off at x = 4 and hits zero at x = 6, which shows, as one might expect, there are very few jobs in trades and services requiring high-level research skills:
The ‘nerdiness ranking’ of each job that you search on PostAc shows where the job falls along this research intensity spectrum. Our machine can categorise huge datasets, which in this system have been supplied by the provider Burning Glass. We hope the wealth of information here will be enlightening and help you think about your job options creatively.