Can language help in the characterization of user behavior? feature engineering experiments with Word2Vec

¹ Among the many significant advances in the area of deep learning, the Natural Language Processing (NLP) space holds a special place. The availability of very large datasets along with the existence of powerful computing environments have created a fascinating environment for researchers. One of the algorithms recently developed is Word2Vec, which enables the creation of embeddings (low-dimensional, meaningful representations of language that can be used for machine learning tasks such as prediction or classification). In this study, we experiment with Word2Vec and apply it to a different domain, i.e., representation of user behavior in information systems. We demonstrate how feature engineering tasks for user behavior characterization can be enriched by the use of NLP concepts.