Recommender systems

In today's digital environment, characterized by an overabundance of information, known as infobesity or the "information deluge", it is clear that human capacities do not allow for exhaustive analysis of the corpus on offer within a platform. Even when using an integrated search engine, relevant results are generally drowned in informational "noise", which prevents them from being found, or at least slows them down. To help the human mind in its selection process, consumer recommendation systems were developed in the last decade of the twentieth century.

A recommendation system is an information filtering tool that helps users make personalized selections from a catalog of items. These systems can be applied in a variety of ways: in social networking sites, in digital marketing with customer relations for online sales, or in personalized services linked to a cultural offering.

After an overview of the fields of application of recommendation engines, the main recommendation strategies are presented from a theoretical and algorithmic point of view. The personalization of these systems can be based on several algorithmic methods, mainly oriented around social aspects and/or the characteristics of the objects handled. This article also highlights the collaborative approach through an example based on open source tools.

With more than 20 years' hindsight on these systems, questions are emerging about ethics, respect for privacy and user trust. As a result, discussions are underway to standardize and provide a legal framework for the phenomenon of recommendation.