In the recruiting and staffing industry, fast and accurate candidate matching is an enormous competitive edge. In many cases, you're in a race to present the right candidate before your competitor identifies the same person in their candidate database. Given the large number of candidates that most recruiters deal with, it's easy to lose track of your best matches.
The Holy Grail of recruiting search applications is accurate, automatic comparison of job vacancies and CVs (Curriculum Vitae). The process is bi-directional and works like this:
- A job description (the document as a whole) is submitted as the search request, and the comparison system returns a short list of the best-qualified candidates from a database of CVs.
- Alternatively, a job seeker (or professional recruiter) submits a CV, and the system returns a list of the most appropriate, currently available vacancies.
<<< Start >>>
<<< End >>>
Automated CV/resume matching approach to improve recruiting efficiency
As with a number of search-related applications, including near-duplicate detection, and document clustering, the usual approach is to extract document vectors for comparison purposes. With CVs and job descriptions, document vectorization is a multi-dimensional task.
Dimensions can include:
- Job titles
- Skills, experience, and qualifications
- Location (how close, geographically, is the candidate to the job)
- Salary range
- Industry sector
A good match will generally find a weighted balance of these vectors.
Tweaking the variables
The need to take multiple dimensions into account necessitates a relevancy benchmarking system. Depending on the circumstances, document vectors can be weighted differently. For example:
- For one job vacancy, proximity to the employer's location may be extremely important, but experience in the role is less so.
- In contrast, another job could require specific skills and experience, and given the salary expectations and scarcity of those skills, candidate location may be less important.
A relevancy benchmarking system can be used to fine-tune the underlying algorithm, in terms of the contribution to the overall match-score made by the various document vectors. Relevancy benchmarking should always be based on human judgment, but much of the process can be automated.
On top of this, users can be given direct control over the weighting of vectors through a dynamic user interface, so that the specific requirements of a vacancy can be reflected in the search results.
At Accenture, we deliver custom recruiting search and match solutions using commercial and open source search engines combined with various Natural Language Processing (NLP) and document understanding techniques. With a powerful search and match system, recruiters can increase in fill rate and recruiter productivity, reduce time-to-fill and job board advertising spend, scale to millions of jobs and candidates, and lower total cost of ownership.