conSultantBERT: Fine-tuned Siamese Sentence-BERT for Matching Jobs and Job Seekers

In this paper we focus on constructing useful embeddings of textual\ninformation in vacancies and resumes, which we aim to incorporate as features\ninto job to job seeker matching models alongside other features. We explain our\ntask where noisy data from parsed resumes, heterogeneous nature of the\ndifferent sources of data, and crosslinguality and multilinguality present\ndomain-specific challenges.\n We address these challenges by fine-tuning a Siamese Sentence-BERT (SBERT)\nmodel, which we call conSultantBERT, using a large-scale, real-world, and high\nquality dataset of over 270,000 resume-vacancy pairs labeled by our staffing\nconsultants. We show how our fine-tuned model significantly outperforms\nunsupervised and supervised baselines that rely on TF-IDF-weighted feature\nvectors and BERT embeddings. In addition, we find our model successfully\nmatches cross-lingual and multilingual textual content.\n