URecommender: An API for Recommendation Systems

• Autor(a) principal: Teruya, Haroldo Shigueaki [UNESP]
• Data de Publicação: 2020
• Outros Autores: Marcal, Ingrid [UNESP], Correia, Ronaldo Celso Messias [UNESP], Garcia, Rogerio Eduardo [UNESP], Eler, Danilo Medeiros [UNESP], Nunes, Joao Osvaldo Rodrigues [UNESP]
• Tipo de documento: Artigo de conferência
• Idioma: por
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.23919/CISTI49556.2020.9141055; http://hdl.handle.net/11449/201992
• topic: cold-start; information filtering; Recommendation systems; recommendations; similarity measure; Web

Recommendation systems are intended to assist users in dealing with information overload by providing a content filtering mechanism. Content filtering is based on the user's preferences and interests. Current recommendation systems suffer from the problem of a lack of initial information about new users. This problem, known as the cold-start problem, is present both in existing systems and in new systems, in which any user is a new user. In addition, web application developers find it difficult to integrate recommendation systems into their applications, having to resort to third-party software or develop the recommendation system from scratch. In this work, URecommender is proposed, an API for web recommendation systems composed of a Middleware and a Framework capable of identifying the textual content of greatest interest to the user and recommending relevant related content. Such identification is done implicitly and based on the user's current behavior, which can solve the cold-start problem. In addition, URecommender gives the developer greater control over the recommendation system that will be integrated into the web application under development. The API was used for the development of a real web application and demonstrated good results in the recommendations generated.