Its yet another job season and you find yourself in the same old tight spot you were in, the last time. Hit that post button and your nightmare starts – only catch, it does not end when you wake up. Sound familiar? A much relatable experience of any and every recruiter out there. Getting hundreds of resumes a day, sifting through each, finding a suitable candidate, shortlisting them, all the while making sure that they are on par with the company’s requirements. This is what prompted our client, among other things, to develop a powerful decision supporting tool, a CV sorting platform using Data science. It allows you to combine the performance of precise sorting with respect to concrete ethical principles. In addition, automation of various processes like Job publication, CV selection, Applicant communication has drastically brought down the workload of recruiters and is time-saving.
• Initially, in the pilot phase, an in-house algorithm was developed by TNP India for resume parsing. They were able to achieve 60% of accuracy for English and not more than 50% for French. Due to the inadequate number of CVs, as well as usage of existing model for layout analysis and NLP, an accurate extraction of data could not be done. A resume, in a completely new format would fail to be acknowledged by the algorithm. Also, the different areas (work experience, education, languages etc.) in a resume, required a different model. Not having as much resources in hand or using pre-existing models made the in-house algorithm lacking in certain areas. Lack or minimum knowledge of French language also proved to be a hurdle.
• In Phase 1, the scoring method was of a basic design and therefore, missed a number of premium resumes due to the inaccuracies in qualification score. Bag of words (Fairjob Dictionary) also had to be enhanced due to the large volume of CVs received.
• Phase 2 made fine adjustments to the final developments that were deployed at this stage. Ontology was an evolving process, as was recruitment profile, throughout this phase. A transparent scoring was one of the more attractive features of this - selecting CVs without being biased, for more than one profile at a time, purely on the basis of talent and skill that are required by a business.
The team at Cloudium and TNP put together a model which was deployed on AWS EC2 platform for a leading bank in France. The following was developed to support the platform:
i. Professions and job profiles modelling – scoring, ranking etc.
ii. An automated system for day-to-day CVs reading and classification.
iii. Data analytics platform to track HR performance indicators and benchmarks.
iv. Development of CV classification Decision Model, whose parameters are transparent and traceable by the client.
v. Rule based classification and ranking.
vi. Resumes are scored along the Completeness and Qualification criteria and displayed in a Decision Matrix with actionable discriminating thresholds.
In addition to that, resume parsing is done by a third-party API – Textkernel.
Python Django has been used for the backend development while Angular 8 has been used for frontend. It’s enabled the localization(L10n) of the website (French and English). Use of visualization tool, Chart.js, helps user understand, have a clear opinion and to arrive at a proper decision. Hosted on AWS platform, it is accessible on laptop/desktop.
Thousands of students graduate each year from universities worldwide. Millions of resumes reach an HR’s desk every day. A few hundreds get placed. But the sad fact is most of them are not even in the right industry. This is what our client solves with their recruitment platform and why they reached out to Cloudium for help. With our ever-driven task force, and top-of-the line technical expertise, Cloudium has a track record of success stories.
Below are a few pointers on what we have done for our our wonderful customer, keeping in mind their requisites-
>> Extracting structured data from the large number of resumes;
>> Scoring and Matching it on the basis of Recruitment profile derived from features contained in the Ontology;
>> Create a premium collection of resumes that aptly describe a company’s business needs; and
>> Making sure that the selected candidate is the right choice.