Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CareerBERT: Matching Resumes to ESCO Jobs in a Shared Embedding Space for Generic Job Recommendations

Julian Rosenberger, Lukas Wolfrum, Sven Weinzierl, Mathias Kraus, Patrick Zschech

TL;DR

CareerBERT introduces a domain-adapted SBERT-based model that learns a shared embedding space for resumes and general job categories by integrating the ESCO taxonomy with real-time EURES job advertisements. It leverages a Siamese SBERT architecture and a multiple negatives ranking objective, trained on ESCO-derived sentence pairs and augmented with a classifier-driven truncation of ads to reduce noise. The system is evaluated through a dual framework: application-grounded evaluation on EURES ads and human-grounded evaluation with HR experts, demonstrating superior performance against traditional embeddings and competitive results with state-of-the-art models, plus robust real-world applicability across diverse resumes. The work contributes a reproducible, modular approach combining taxonomic knowledge with live job data, offering practical tools for career counselors and a path toward scalable, language-wide job matching with bias-aware, explainable potential.

Abstract

The rapidly evolving labor market, driven by technological advancements and economic shifts, presents significant challenges for traditional job matching and consultation services. In response, we introduce an advanced support tool for career counselors and job seekers based on CareerBERT, a novel approach that leverages the power of unstructured textual data sources, such as resumes, to provide more accurate and comprehensive job recommendations. In contrast to previous approaches that primarily focus on job recommendations based on a fixed set of concrete job advertisements, our approach involves the creation of a corpus that combines data from the European Skills, Competences, and Occupations (ESCO) taxonomy and EURopean Employment Services (EURES) job advertisements, ensuring an up-to-date and well-defined representation of general job titles in the labor market. Our two-step evaluation approach, consisting of an application-grounded evaluation using EURES job advertisements and a human-grounded evaluation using real-world resumes and Human Resources (HR) expert feedback, provides a comprehensive assessment of CareerBERT's performance. Our experimental results demonstrate that CareerBERT outperforms both traditional and state-of-the-art embedding approaches while showing robust effectiveness in human expert evaluations. These results confirm the effectiveness of CareerBERT in supporting career consultants by generating relevant job recommendations based on resumes, ultimately enhancing the efficiency of job consultations and expanding the perspectives of job seekers. This research contributes to the field of NLP and job recommendation systems, offering valuable insights for both researchers and practitioners in the domain of career consulting and job matching.

CareerBERT: Matching Resumes to ESCO Jobs in a Shared Embedding Space for Generic Job Recommendations

TL;DR

CareerBERT introduces a domain-adapted SBERT-based model that learns a shared embedding space for resumes and general job categories by integrating the ESCO taxonomy with real-time EURES job advertisements. It leverages a Siamese SBERT architecture and a multiple negatives ranking objective, trained on ESCO-derived sentence pairs and augmented with a classifier-driven truncation of ads to reduce noise. The system is evaluated through a dual framework: application-grounded evaluation on EURES ads and human-grounded evaluation with HR experts, demonstrating superior performance against traditional embeddings and competitive results with state-of-the-art models, plus robust real-world applicability across diverse resumes. The work contributes a reproducible, modular approach combining taxonomic knowledge with live job data, offering practical tools for career counselors and a path toward scalable, language-wide job matching with bias-aware, explainable potential.

Abstract

The rapidly evolving labor market, driven by technological advancements and economic shifts, presents significant challenges for traditional job matching and consultation services. In response, we introduce an advanced support tool for career counselors and job seekers based on CareerBERT, a novel approach that leverages the power of unstructured textual data sources, such as resumes, to provide more accurate and comprehensive job recommendations. In contrast to previous approaches that primarily focus on job recommendations based on a fixed set of concrete job advertisements, our approach involves the creation of a corpus that combines data from the European Skills, Competences, and Occupations (ESCO) taxonomy and EURopean Employment Services (EURES) job advertisements, ensuring an up-to-date and well-defined representation of general job titles in the labor market. Our two-step evaluation approach, consisting of an application-grounded evaluation using EURES job advertisements and a human-grounded evaluation using real-world resumes and Human Resources (HR) expert feedback, provides a comprehensive assessment of CareerBERT's performance. Our experimental results demonstrate that CareerBERT outperforms both traditional and state-of-the-art embedding approaches while showing robust effectiveness in human expert evaluations. These results confirm the effectiveness of CareerBERT in supporting career consultants by generating relevant job recommendations based on resumes, ultimately enhancing the efficiency of job consultations and expanding the perspectives of job seekers. This research contributes to the field of NLP and job recommendation systems, offering valuable insights for both researchers and practitioners in the domain of career consulting and job matching.

Paper Structure

This paper contains 25 sections, 4 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Foundations and Related Work
  3. NLP Techniques for Job Matching
  4. Adapting Language Models for the Job Domain
  5. Advances in NLP-Driven Job Matching
  6. Development of Job Recommendation System with CareerBERT
  7. Overview of the Job Recommendation System based on CareerBERT
  8. Preprocessing of Job Data
  9. Incorporating EURES Job Advertisements
  10. Job Advertisement Centroids and Job Centroids
  11. Results
  12. Experimental Setup
  13. Base Model Selection
  14. Training Data Selection
  15. Training Process
  16. ...and 10 more sections

Figures (4)

  • Figure 1: Workflow of the proposed job recommendation system with CareerBERT at its core. A user's resume is transformed into a high-dimensional embedding (Resume Vector) through CareerBERT's embedding process. Similarly, jobs are converted into embeddings (Job Vectors) in the same space. The system then finds suitable job recommendations by measuring the cosine similarity between the vectors, where smaller angles ($\theta$) indicate higher similarity.
  • Figure 2: Overview of CareerBERT's training process. Starting with either GBERT or jobGBERT (a domain-adapted model for Human Resources (HR) gnehm2022evaluation) as base models, we explored variants with and without task-adaptive pre-training using Transformer-based Denoising AutoEncoder (TSDAE) before final training on ESCO sentence pairs.
  • Figure 3: Visualization of human-grounded evaluation. Resumes are processed by CareerBERTJG using the best-performing jobGBERT model and job centroid embeddings
  • Figure 4: CareerBERT application interface deployed using Streamlit. Users can upload their resumes and receive personalized job recommendations based on the CareerBERT model. The interface allows users to select a model, upload a resume, and customize the text for better job matching.