Active Hackathon

How LinkedIn Is Using Deep Learning To Increase Hiring Efficiency Amid Recession

When it comes to job hunting, there is no other place than the largest professional and employment-oriented service platform LinkedIn. With hosting over 20 million active job postings, the largest hiring marketplace has been continuously developing its platform with the help of using intelligent models to optimise various processes such as job postings, job recommendations, handling abusive contents and much more.

To the latest, the developers at LinkedIn recently unveiled a new deep learning model known as Job2Questions. Using it, recruiters can ask screening questions online to filter qualified candidates easily. The professional social networking platform has many conventional on-demand services and state-of-the-art machine learning models that help millennials and professionals in the global workforce. 


Sign up for your weekly dose of what's up in emerging technology.

The developers stated that the primary goal is to match jobs with qualified applicants and improve hiring efficiency while reducing the requirement of manually screening each applicant. 

The Deep Learning Model

In order to solve the issue of generating screening questions, the researchers at LinkedIn developed a two-stage deep learning model known as Job2Questions. In this method, a deep learning model is applied to detect intent from the text description and rank the detected intents by their importance based on other contextual features.

Job2Questions is a Screening Question Generation (SQG) model that automatically generates structured SQs for a given job posting. According to the developers, this AI product is new compared to the earlier products and has no historical data. The researchers employed deep transfer learning to train complex models with limited training data.

Why This Model

Most of the time, interviewing applicants for a job profile is costly and inefficient. A traditional hiring method follows a number of steps such as screening the applicants in the pool by their profile and conducting additional phone screenings before sending out interview invites. This process is not only time consuming but also ends up in finding out such applicants who have missing basic qualifications such as work authorisation/visa, minimum years of experience, or degree requirements.

This is where Job2Questions comes into play. This new model addresses such issues, ease the manual workload by estimating person-job fit automatically. It is a Screening Question (SQ) -based online screening product for LinkedIn, which works by asking job-specific questions to applicants and assessing them using the answers they provide when applying for the job.


The researchers mentioned two significant product challenges for designing a successful SQ-based online screening product:-

  • The product should provide an easy way to add SQs to jobs, since asking the recruiters to reformulate their job postings into SQs manually is a waste of time and human resources.
  • SQs should help recruiters identify qualified applicants quickly. A job requirement may have many different expressions, using unstructured text questions to present SQs will be tough for recruiters or AI models to interpret the intent of the SQ.


This new AI model has a number of benefits, as mentioned below:

  • Job2Questions model helps recruiters to find qualified applicants, and job seekers to find openings they are qualified for
  • Job2Questions model is cost-effective and addresses hiring inefficiency
  • The model has been created to solve hiring issues while making it easy
  • Job2Questions eases the manual workload by estimating person-job fit automatically

Contributions To This Research 

  • According to the researchers, this is the first work on the Screening Question Generation (SQG) task, which generates structured screening questions to help assess job applicants
  • The researchers proposed and deployed the first SQG model, Job2Questions in production to help millions of jobs finding qualified applicants and assist hundreds of millions of members in identifying qualified jobs
  • During an offline evaluation, the proposed Job2Questions model improved the AUROC of both template classification and question ranking by 178% and 27.4%, respectively
  • Job2Questions significantly improved the online SQ suggestion quality by +53.10% acceptance rate and +22.17% job coverage

Wrapping Up

According to the developers, the online A/B testing showed +53.10% screening question suggestion acceptance rate, +22.17% job coverage, +190% recruiter applicant interaction, and +11 Net Promoter Score.

In one of our articles, we discussed how the recommendation system of LinkedIn is generating the perfect job match. Amid the pandemic, the platform has also announced to offer free job postings to accelerate hiring for critical roles during the COVID-19 crisis. This initiative has been launched to promote urgent jobs and recruiting resources to organisations in critical need such as healthcare, supermarket, warehousing and freight delivery, including disaster relief nonprofits to help them find talent quickly.

Read the paper here.

More Great AIM Stories

Ambika Choudhury
A Technical Journalist who loves writing about Machine Learning and Artificial Intelligence. A lover of music, writing and learning something out of the box.

Our Upcoming Events

Conference, in-person (Bangalore)
Machine Learning Developers Summit (MLDS) 2023
19-20th Jan, 2023

Conference, in-person (Bangalore)
Data Engineering Summit (DES) 2023
27-28th Apr, 2023

Conference, in-person (Bangalore)
MachineCon 2023
23rd Jun, 2023

3 Ways to Join our Community

Discord Server

Stay Connected with a larger ecosystem of data science and ML Professionals

Telegram Channel

Discover special offers, top stories, upcoming events, and more.

Subscribe to our newsletter

Get the latest updates from AIM