JobScan: Automated CV-Vacancy Matching and Improved Search in a Vacancy Database

JobScan: Automated CV-Vacancy Matching and Improved Search in a Vacancy Database

Gupta, A.

Rothkrantz, L.J.M. (mentor)

Electrical Engineering, Mathematics and Computer Science

Man-Machine Interaction Group

2012-09-01

In the last few years there appears to be a paradigm shift in the job recruitment field, from physical contact between recruiters and job seekers to e-recruiting, with online job boards providing the meeting place for both parties. As job boards have gained in popularity so have the number of vacancies and curricula vitae being posted on these sites. Whilst this wealth of online vacancies and CVs provides great opportunities to recruiters and job seekers alike, locating relevant information on the job boards has become an increasingly tedious task. This thesis describes an attempt to alleviate this situation by creating a web application that aids both recruiters and job seekers. Recruiters are offered the possibility to match a CV to all vacancies in the database, providing them with possible leads. Job seekers will also appreciate this functionality as it provides them with job suggestions and saves them time searching through many vacancies. Job seekers are also provided with improved means to search an online vacancy database, through offering them advanced (boolean) search capabilities and improved result ranking. In both cases the effort required to find suitable matches is reduced. A local vacancy database was created by retrieving vacancies from a number of popular Dutch job boards. Data extractors were implemented to automatically extract features such as the function title, job requirements text and required education, skills and competences from the retrieved vacancies, after which the vacancies were clustered on the requirements text using the cosine similarity measure and k-means. The cosine similarity measure was also used as part of the CV vacancy matching algorithm and to improve result ranking when querying the vacancy database. Evaluation of the working prototype demonstrated that whilst some improvement to the data extractors is desirable, the prototype was able to show that improved result ranking and relevant CV-vacancy matches are achievable when CVs and vacancies are of high quality.

http://resolver.tudelft.nl/uuid:d7bdf7d3-3431-4440-b6de-afa368dd0da0

Student theses

master thesis

(c) 2012 Gupta, A.