Saturday 10:30 AM–11:15 AM in C11

Intelligent recruitment using NLP and machine learning to identify the most suitable candidates

Victor Robin

Audience level:
Novice

Description

It is very common to get 100s of applicants for a role, which is becoming difficult for recruiters to shortlist the most relevant candidates. We present here a solution developed at dunnhumby that leverages Natural Language Processing to assist our recruitment team. By simply providing a reference resume, our tool helps identifying similar candidates, helping us focus on the right candidates.

Abstract

The talk presents a real-life example of application of machine learning and natural language processing for the industry.

For every role we advertise, we get 100s if not 1000s of applicants. Shortlisting the most relevant candidates becomes therefore very difficult, and no recruitment team has the capacity and time to go through all the CV of the candidates who applied for a role. It also increases the chances to miss the ideal candidate. We present here a solution developed internally that leverages natural language processing to shortlist the most relevant candidates.

Very often, the hiring manager provides the recruitment team with a reference CV of the ideal profile he or she is looking for. The way we have designed the tool is as follows: The text from all the CVs is extracted and converted into a vector format using Word2Vec. This generates a vector representation of the candidate’s CV, which are then compared using cosine-similarities with the vector of the reference CV. This returns the CV that are the most similar to our reference profile.

The methodology has proven to give good results, particularly efficient at shortlisting the most promising candidates. It still does not replace an actual interview, but it has the potential of greatly reducing the hiring time and ensuring only the most promising candidates are interviewed, saving time for both the manager and the candidates.

Subscribe to Receive PyData Updates