August 14, 2022

ptemplates

Born to play

An AI-based Talent Acquisition and Benchmarking for Job

The recruitment industry spends a lot of person-hours on screening profiles. Not only does it make the whole recruitment process long, slow and cumbersome. Besides, recruiters’ bias can also affect the quality of shortlisting and hiring. 

Talent acquisition is time-consuming process.

Talent acquisition is time-consuming process. Image credit: Pxhere, CC0 Public Domain

In a new article recently published on arXiv.org authors analyze possibility of using AI-based talent acquisition and benchmarking platform for employment purposes. This research paper forms the basis of the following text.

Importance of this research

The average time to fill an open position is 42 days, and employers spend approximately $4,129 to close any job position. Even then, if the recruiter selected the right candidate isn’t easy to ascertain. AI can make the overall Recruitment process more efficient. This research paper has proposed a way to identify and separate more relevant resumes from less relevant ones. A relevance score between job-seekers and employers based on skill match and culture match leads to AI-assisted screening. This research can help recruiters make unbiased decisions and help recruiters save money and time in the recruitment process.

The researchers have also credited the existing AI-driven talent solutions in their research paper.  

Image credit: arXiv:2009.09088 [cs.CY]

Scope of the research

To keep things simple, the researchers have limited their project to solve the problem in the computer science industry only

HAY is a commonly used benchmarking criterion for jobs in the recruitment industry

  • Know-How: Measures the range of skills as described in the Job post
  • Problem Solving: Measures the degree of complexity associated with the Job
  • Accountability: Measures the level of responsibility required for the role. 

Implementation of the project

Technical skills and cultural fit criteria were considered for mapping CVs and the job postings. 

Measuring Technical Fit:

Key points considered for technical fitment are as below.

  • Same OR-related words need to be accounted for better matching. For example, ML and Machine Learning are the same. 
  • Some words could have a parent-child relationship. For example, if the classifier extracts the term “machine learning” it will also infer “artificial intelligence”. 
  • The frequency of specific skill keywords (or closely related) was considered while establishing relevance between the job-post and resume.
  • Education was also considered for matching job-posts and resumes.

Measuring Cultural Fit:

  • The Job-post identifies Organizational culture, measures it based on the below bi-polar criteria, and maps with the candidate to establish their cultural fit. 

Image credit: arXiv:2009.09088 [cs.CY]

Use cases

The below two use-cases were used to match the job-post and resumes

  • ManyToOne matching: This benchmarks different CVs for a specific job-post and assigns a matching score to them. 
  • OneToOne matching: This explores the level of correspondence between a particular job-post and resume.

Data Set

The algorithm was used on five different resumes and considered for a Data Science Internship

Result

The proposed algorithm was able to identify relevant resumes for a specific Job and also establish the relative of correspondence between various resumes and job-post using ManyToOne matching and OneToOne matching, respectively. 

Conclusion

In the words of the researchers

About the process, a clear algorithm for recommendation could be implemented. Possible models for the CV parsing and recommendation processes are variate, not a complete approach can be found in research. We implemented a divide and conquer methodology for the model. We can approach each problem and solve each one with the best tools such as ontologies, embeddings, direct match, expert evaluation, machine learning. Develop an algorithm for the whole process according to the existence or not of data. About the product, it would reduce time in the recruiting process, save money, invest recruiters in more productive activities in order to increase retention and productivity of the team and decrease recruiters bias. Besides, it would encourage best candidatesfitting on the organization, which would increase the company value as a consequence.

Source: Rudresh Mishra, Ricardo Rodriguez and Valentin Portillo,  “An AI Based Talent Acquisition and Benchmarking for Job”.