Recommender system for sports articles based on Arabic opinions polarity detection with a hybrid approach RSS-SVM

In this paper, an Arabic recommender system based on opinion analysis and polarity detection is proposed. Unfortunately, working with Arabic adds more difficulties than the other languages, because it implies the solving of different types of problems such as the diversity of dialects, Al hamza, the ambiguity, etc. These sorts of applications produce data with a large number of features, while the number of samples is limited. The large number of features compared to the number of samples causes over-training when proper measures are not taken. The aim of this work is to combine both the random sub space method and support vector machine classifier in order to avoid over fitting creating by the used of all features and beneficiate from proven SVM classifier performances. The main steps of this study are based primarily on articles collection, Statistical features extraction, opinions polarity detection and then generating the recommendations by the proposed hybrid approach. Experiments results based on 1000 comments collected from Algerian sports web sites are very encouraging.

I. Introduction

The goal of a Recommender System is to generate meaningful recommendations to users for items or products that might interest them. This new area of research is becoming more and more important mainly due to the growth of social media; most of built resources and systems are intended to English or other European languages. Despite the fact that Chinese, Arabic and Spanish are currently among the top ten languages most used on the Internet, there are very few resources for recommending users in these languages.