Machine Learning Model for Personalizing Online Arabic Journalism

The paper discusses a model of generating dynamic profile for Arabic News Users, capturing user preference by analyzing his review of historical news, and recommend him news as soon as he creates account on News Mobile App, Preference is calculated based on article main keywords score, which is extracted from article headline & body as NLP (Natural Language Processing), when user reads an article, its keywords are calculated with rate of interest to his profile. Machine Learning techniques are used in the proposed model to recommend user the relevant news to his preferences and provide him personalization. The model used hybrid filtering techniques to recommend user suitable articles to his preferences, as Content-Based, Collaborative, and Demographic filtering techniques with KNN (K-nearest neighborhood). The model combined between those techniques to enhance the recommendation process, after recommendation happened, that the model tracks User behavior with the recommended articles, whether he reviewed it or not, and the actions he did on the article page to calculate his rate of interest, then dynamically updates his profile in real time with interested keywords score, thus By having User profile and defined preference, the model can help Arabic news publisher to classify users into segments, and track changes in their opinion and inclination, using observation method of read news from different user segments, and which articles attract them, thus it leads publishers to visualize their data and raise their profitability, and to follow the international trend in e-journalism industry to be a data driven organization.