• document embeddings
• document similarity
• job matching
• ranking

Matching job offers and candidates is a challenging task that can be solved in different ways.
This work proposes an experimental study oriented to overcome classical Boolean Keyword Search matching systems, often based on arbitrary weighting scheme. This improvement can be achieved by drawing concepts from Natural Language Processing and Text Mining field area, in particular for document-to-vector transformation. Document embeddings production algorithms are combined with
common machine learning algorithms to design and build an automated tool able to produce ranked lists of matching candidates for a certain job offer.
Advantages and drawbacks about these more intelligent models are listed, after having verified that this paradigm
adds an undeniable improvement, especially when used in synergy with pre-existing keyword-search focused systems. Starting from this experiments, an automated tool to produce candidates ranking has been designed and implemented.