Magnet.me  -  The smart network where hbo and wo students find their internship and first job.

The smart network where hbo and wo students find their internship and first job.

Data Science thesis (MSc): Stemming Dutch job titles

Posted 1 Feb 2024
Share:
Work experience
0 to 1 years
Full-time / part-time
Full-time
Job function
Degree level
Required languages
English (Good)
Dutch (Fluent)
Start date
1 April 2024

Your career starts on Magnet.me

Create a profile and receive smart job recommendations based on your liked jobs.

Introduction

In the contemporary job market, efficient talent acquisition relies heavily on precise matching of job seekers with relevant job opportunities. One crucial aspect of this process is the accurate analysis of job titles, as they play a pivotal role in conveying the nature and requirements of a position. This thesis proposes the design and implementation of a specialized stemmer for job titles in Dutch, with the aim of enhancing semantic matching capabilities within the recruitment platform at Magnet.me.

Objective

The primary objective of this thesis is to create a robust and context-aware stemmer tailored to the unique linguistic nuances of job titles in the Dutch job market. By developing an algorithm that intelligently identifies and processes variations of job titles, we aim to improve the accuracy of job classification at Magnet.me. The stemmer will serve as a vital component in classifying jobs, which in turn is a key component of our search functionality and recommendation systems.

(Proposed) Research Questions

  • What linguistic challenges and variations exist within job titles in the Netherlands that impact the accuracy of semantic matching?
  • How can a stemmer be designed to effectively address these challenges and accurately capture the semantic essence of Dutch job titles?
  • What metrics can be employed to evaluate the performance and effectiveness of the proposed stemmer in enhancing job title classification at Magnet.me?

Methodology

The research will adopt a mixed-methods approach, combining qualitative analysis of Dutch job titles, linguistic patterns, and user feedback with quantitative assessments of the stemmer's performance. Natural Language Processing (NLP) techniques and machine learning algorithms will be explored to develop a context-aware stemmer that adapts to the dynamic nature of job titles.

Significance

The successful implementation of a Dutch stemmer for job titles holds immense significance for Magnet.me, as it directly contributes to the platform's mission of facilitating meaningful connections between job seekers and employers. The improved semantic matching will enhance user experience, improve recommendation quality, reduce search friction, and ultimately foster more efficient and accurate job placements.

Expected Outcomes

  • A comprehensive understanding of linguistic variations in Dutch job titles.
  • The development and implementation of a specialized stemmer for job titles in Dutch.
  • Quantifiable improvements in the accuracy of semantic matching at Magnet.me, as evidenced by user engagement and feedback.

Conclusion

This thesis seeks to bridge the gap between language processing technology and the intricacies of Dutch job titles, offering a tailored solution to optimize semantic matching within the Magnet.me platform. The outcomes of this research have the potential to significantly elevate the effectiveness of talent acquisition processes, benefiting both job seekers and employers.

Who are we?

Magnet.me is the #1 career network for students, graduates and employers in the Netherlands.

Magnet.me uses ML to empower every student to build a network of interested employers and to discover jobs based on their profile and preferences. More than 300,000 students are on the network to kickstart their careers.

On the employer side, 5,000+ companies - from multinationals like Unilever, Heineken and McKinsey to startups and scale ups - use Magnet.me to establish meaningful connections with young talent and to fill their internships and jobs without hassle.

We are a small team of ~25 Magneteers. You will be part of the Product Team, where we develop our own platform; including our Machine Learning solutions. You will be supervised by Alex Walterbos, MSc (TUDelft) and Rogier Slag MSc (TUDelft).

You

  • are a Master student with a relevant study background (Data Science or comparable)
  • are looking for a graduate internship in Recommender Systems and an interesting thesis topic
  • want to gain practical experience in a dynamic company
  • want your work to have a real impact!

Practical info

  • Magnet.me works hybrid, so working from home is possible
  • Start- and end dates will be determined in collaboration
  • Magnet.me's recommender systems are built in Python (with sklearn, pandas) and use the Google Cloud Platform.
  • Earliest start date is 2024-04-01, but a later kickoff is possible.

Interested? We are!

Apply below to get in touch!

Magnet.me is the #1 career network for students, graduates and employers in the Netherlands. Magnet.me uses AI to empower every student to build a network of interested employers and to discover jobs based on their profile and preferences. Companies use Magnet.me to establish meaningful connections with young talent and to fill their internships and jobs without hassle. More than 300,000 students and 5,000 companies, from multinationals to startups, are currently on the network.

IT
Rotterdam
Active in 2 countries
33 employees
70% men - 30% women
Average age is 27 years

What employees are saying

Bette Donker

Data Analyst

Bette Donker

I really like that everyone at Magnet.me is very open and works hard towards a common ambitious goal. As a Data Analyst, I monitor our marketplace to provide insightful data for informed decisions. This role challenges me both strategically and technically, which stimulates me to keep on learning!

Mike Rovroy

Product Designer UX/UI

Mike Rovroy

What I love about Magnet.me, is its versatility. We have a great mix of creative marketeers, a helpful support squad, typical Delft developers, and customer-oriented sales people. Everyone adds something, especially to the great atmosphere during Friday afternoon drinks.