Is it Required? Ranking the Skills Required for a Job-Title
This addresses the need for automated skill ranking in job markets, but it is incremental as it builds on existing language models and weak supervision techniques.
The paper tackles the problem of ranking skills required for job titles by training a LaBSE model with weak supervision, showing it can learn skill importance and perform well across languages, with concrete improvements from using Inverse Document Frequency to boost specialized skills.
In this paper, we describe our method for ranking the skills required for a given job title. Our analysis shows that important/relevant skills appear more frequently in similar job titles. We train a Language-agnostic BERT Sentence Encoder (LaBSE) model to predict the importance of the skills using weak supervision. We show the model can learn the importance of skills and perform well in other languages. Furthermore, we show how the Inverse Document Frequency factor of skill boosts the specialised skills.