<![CDATA[One of the main forms of information and entertainment presentation is in the form of on-demand audio content or often referred as podcast. The problem met is the difficulty of users to find podcasts that match their preferences among hundreds of thousands of podcasts that are now available on the internet. Besides the number of podcasts, the incompletion of podcasts metadata is one of the problems in developing a recommendation system in general.This system is designed by utilizing N-Gram and Term Frequency to generate queries from the title and description of a podcast which will then be used to find other podcasts that are similar by utilizing Cosine Similarity calculations. The testing phase is done through by testing the system using nDCG calculations to determine the relevance level of the recommendation results. From the results, the average value of the relevance level of podcast recommendations is 53.8% and the best average f1-score of the 10 best categories is 14%.]]>
Copyrights © 2020
