Content-based Filtering Model for Recommendation of Indonesian Legal Article Study Case of Klinik Hukumonline

Recommender system help users in getting the relevant information. For example, it may advise users on a related topic of article, or suggest complementary items to purchase. Application of recommender system is commonly found in commerce sites. Hukumonline is one of media companies and law services. The website consists of various legal content, such as news, consultation articles (they call it Klinik), data repository, event information, and journal (currently beta version). At present, Hukumonline site conduct a manual recommendation system, annotated by the content team as their daily routines. This paper describes our experiment of Content-based filtering (CBF) model for recommendation of Bahasa Indonesia to users of Hukumonline Klinik article on the site. So the manual and labor intensive process can be reduced. For this purpose, we use supervised learning method. Starting with data collection of 3,700 articles spread in 15 categories, followed by preprocessing, vector space representation, and the learning phase, which we experiment on K-Nearest Neighbor algorithm with cosine similarity as the distance metric. We use number hyperparameter K of 17. We separated ten percent of articles for our test data. For evaluation, we use K-Fold Cross Validation with K of 10. Our model generate accuracy rate of 0.75, precision rate of 0.76, recall value of 0.77, and F-Measure of 0.75.