This paper is about the adaptive and personalized justification of the recommenders collaborative filtering system using notices. A method to justify recommendations based on item reviews and the user profile interest is suggested. The reviews with a positive sentiment have been first kept through the sentiment analysis expressed on the reviews and then second, selected potential reviews are candidates for the justification of items. To identify reviews candidates, the frequency calculation of user profile interest terms in the reviews has been done through the TF-IDF weighting method. In order to manage our reviews, an algorithm removing negative sentiment reviews is proposed. However, to test the method, recommendation data already made on a collaborative filtering recommendation system using notices and reviews made on Coursera courses have been used. The data used included 112 recommendations and 11 users. The implementation shows that the inference is constantly evolving and increasingly adapted to the user’s profile.

Personalization, Justified Recommendations, Users Profile, Interest, Reviews