Computers use natural language processing to learn human language (to the extent that they can). Augmented writing uses natural language processing and probability to offer guidance on spelling, grammar, and (in a limited way) word choice. But what does language analytics do? And, more specifically, what is language analytics for job posts?
Language analytics for job posts uses natural language processing, behavioral science, and data science to offer real-time guidance on language and content in job posts. Unlike augmented writing, which relies on a single factor (probability), language analytics uses a number of factors. They include titling, language clarity, requirements, bias, content (i.e., a diversity statement), and many more.
Language analytics for job posts, briefly
Language analytics for job posts determines the statistical probability of success for a job post (i.e., how well it attracts more qualified, diverse candidate pools) based on the success of other real-world job posts. It then makes recommendations for increasing that probability.
As a starting point, it’s hard to offer good guidance on job posts without first determining what an inclusive, effective job post looks like. Doing that requires analyzing millions of job posts and figuring out what makes the successful ones tick. Effective job posts include titles that candidates can find, clear requirements, and inclusive language that welcomes all to apply.
But their success also depends on factors such as a company’s compensation levels, remote-work policies, and commitment to diversity, equity, and inclusion.
For example, does an industry-standard job title improve a job post (e.g., ‘Accountant’ as opposed to ‘Numbers Ninja’)? Yes, because candidates use common titles in their job board searches (i.e., ‘Accountant’ will show up, ‘Numbers Ninja’ may not). Does an accommodation statement make a job post better? Yes, because it reassures disabled job seekers that an organization will take steps to ensure their ability to perform in the job.
Language analytics looks at the impact that language and content have on the effectiveness of messaging. Basically, it measures how well the language and content in your job post works. It can do things like identify exclusionary language (i.e., sexist, racist, ageist), recognize wordiness, and even understand concepts or sentiments even when they’re expressed using different phrases (i.e., euphemisms). (That last one is crucial.)
Why language analytics for job posts
A job post is a particular type of writing. It’s got its own lingo (e.g., SQL, Java, SEO) and its own rules (e.g., sentence fragments such as ‘approves cash disbursements’ are okay). And, again, its success depends on a host of factors.
Let’s say, for example, that you have 100 job posts from 100 different companies and 10 of them attracted more women than the rest. Let’s say those 10 job posts all contained the word ‘collaborative’ in them. Does that mean including the word ‘collaborative’ means a job post will perform well with women?
Possibly, but it’s just one factor. And you can’t use a single factor to accurately predict an outcome that’s determined by many factors (e.g., women applying to your job). Without analyzing the other variables, all you know for certain is that these job posts all contained the word ‘collaborative,’ which could be a coincidence. For instance, it could be that only these 10 jobs of the 100 included diversity statements and a benefits section.
Another thing is that there are many ways to say the same thing. Machine learning, while great at spotting spelling errors, isn’t good at understanding the various ways to communicate an idea. An augmented writing tool, for example, doesn’t understand that ‘innovate’ and ‘think outside the box’ can be synonymous. Language analytics for job posts, on the other hand, does. Which means it can make helpful suggestions for alternatives.