Learner comments-based Recommendation system
Manar Joundy Hazar, Mounir Zrigui, Mohsen Maraoui
Abstract
Traditional techniques of recommender systems presented low accuracy performance because it suffer from sparsity problem hence. In this work we build a recommender model based on free-text reviews written online by user. The model built on two method recommendation and sentiment analysis. Our recommendation model first predicts new rating from user reviews then detect user requirements and interest by differences analysis between a new rating prediction and original rating. Presented model take in account both explicit original rating and implicit user sentiments, hence that will be roll key to address data sparsity problem. In this model we try to extract feature and opinion by sentiment analysis of text user comments and reviews to get more accurate rating calculation from student reviews and comments written in English. This model tested and trained on online learner comments and reviews on video learning of three courses on coursera learning platform to produce a video learning recommendation system and to validate the accuracy and efficiency of proposed model.