Follow Africa: Building an African News Recommender Systems

Abstract

Recommender systems have become an important component of Web media. From VoD providers (Netflix, Amazon Video) (https://goo.gl/g3G1ys) to news websites (Yahoo! News, CNN) (https://goo.gl/yA2NB6), users have become accustomed to personalized content. However, news recommendation differs from traditional recommendation due to the short lifetime of news. Indeed, News is particularly characterized by a short time span during which they are relevant. Therefore, in addition, to suggest suited news to users, news recommender systems (NRS) have to deal with news recency in order to avoid recommending already read content somewhere else. In most of the cases, NRS implicitly collect users’ click history and readings to build topic-based user profiles. News websites generally integrate some keywords into the news articles which sum up their content. But this is not always the case for African news websites. In this paper, we present Follow Africa, an African news recommender system. We introduce a recency-based recommendation model which also takes account of users’ previous readings. We show the effectiveness of our proposal through the results we obtain in a month-lasted online experiment with more than one hundred users.