Create Homogeneous Consistent Features from Unstructured and Structured Data Sets Comprising Vacancies, Resumes, Application Forms, Test Scores and Shared LinkedIn Profiles

Project Description

Background

We are a provider of eRecruitment technology which is used by our clients to manage the workflow of recruiting new hires including the following steps: posting vacancies, providing online application forms, integration of recruitment tests, communication with candidates etc.

We host online job application processes for our clients, where clients create vacancies and applicants typically complete a structured application form comprising contact details, education scores and possibly work experience, applicants can also upload their resume/CV as part of their job application. Some candidates also provide the url to their LinkedIn profile and through the LinkedIn API allow us to access their details.

Our clients’ HR managers and recruiters use the candidates’ education, work history and past leadership achievements to select candidates for job interviews.

For recruitment into graduate level jobs, e.g. graduate intern, analyst and associate roles, we have developed a target list of important achievements and leadership positions (classified into 19 categories) which are deemed prestigious or important by recruiters.  Some target terms are generic, others are specific to individual universities or countries.  This target list is constructed using regex, for search purposes, and each achievement has an associated score value.  For example “President of the University Student Finance Society” or “All-American Basketball team member” might be awarded 10 points.

We have a large database, approx. 5 million, of existing resumes/CVs which are stored in pdf format, a similar number of structured application forms, and a smaller number of shared LinkedIn profiles.  Additionally we have 10,000’s of vacancies consisting of structured and unstructured data.

Goals

The goal is to develop a solution which irrespective of the input

•    A structured application form
 
•    A pdf based resume or
  
•    A LinkedIn profile (both a shared profile or a link)
  
•    A wide variety of vacancy templates (consisting of structured and unstructured data)
  
identifies candidates achievements, work experience and education and vacancy requirements and outputs them in a uniform structured format in which equivalent items (with differing descriptions) are recognized as such.

The resulting features will then be used in machine learning algorithms to predict successful job applicants.

Deliverables

The solution would need to be developed as a service with an API to work with our proprietary system.

Skills required
Data management
Data cleaning
NLP
Text scrapping expertise
Creating APIs

Milestones/deadline

We are looking for a working solution that we can implement into our systems during Q3 2017.

Note that we have posted three related projects and are willing to work with one supplier on all three, or with separate suppliers according to expertise and interest.

The projects are:

•    “Create consistent work experiences from shared Linkedin profiles, resumes and structured application forms “

•   “Create homogeneous consistent features from unstructured and structured data sets comprising vacancies, resumes, application forms, test scores and shared LinkedIn profiles”

•    “Use machine learning to predict successful hires from homogeneous features collected from vacancies, resumes, and application forms”