Candidate Summary and Suggested Skills Tool

Published May 28, 2020 by Scott Hirsch

This is the third of five posts about machine learning at TalentMarketplace. Take a quick read of the introduction blog here.

Mukut, a Postdoctoral Fellow at TRIUMF, completed his Masters of Physics and Post Doctorate of Philosophy in Physics. While he leads experimental efforts to perform atomic parity nonconservation measurements in francium atoms (quite a mouthful) by day, Mukut spends his evenings submerged in data science and machine learning.

Mukut worked with TalentMarketplace and embarked on the journey of creating the Candidate Summary and Suggested Skills Summary.

The problem Mukut set out to solve was seemingly simple: when searching terms on TalentMarketplace, the search would only result in an exact match. For example, if an employer searched for a “Business Analyst”, they would miss all the consultants, functional analysts, and other candidates who perform the same function, but with a different title. To save time for employers and increase the success rate of candidates, Mukut set out to find an automated process to find similarities between candidate resumes.

In order to make this happen, Mukut was in need of a model which took into account the similarities between words within a candidate profile. An automated process needed to be created to match word to word skills.

In order for search results to return sufficiently related candidates, Mukut turned to Technical Natural Language Processing. Mukut built a set of algorithms with Word2vec to train neural networks to understand the similarities between words (specifically skills) within a candidate profile.This would then find similarity between different words or skills, even if they do not exactly match.

With the combination of Technical Natural Language Processing, Word2vec, and tech strength, this engine can deliver value on both sides of the market by:

  1. Providing employers with automated summaries of candidates to speed up their review times.
  2. Suggests skills (keywords) to candidates - skills the algorithm expects them to have based on their experience, but ones they did not mention - ones which are proven to help them get an interview.

If you’re seeing a theme here, you’re right: many of these algorithms have applications for both candidates and employers to match with each other. It’s particularly exciting (for me, anyway) how this project helps candidates and employers communicate more seamlessly with each-other. As most traditional recruiters will tell you, this is a big part of our job, and it’s cool to see how machine learning can make us more effective in doing that job.