Abstract: Recommend personalized contents to users be strongly related to their profiles which reflected on the accuracy of recommendation system. In this study, the user profile has been built based on content and context of his preferences. The improved user profile is constructed by finding the correlation among the content and context of items. Given the user and item profiles, Jaccard’s coefficient is applied to compute the similarity between pair of profiles, hence, k-nearest neighbors of items have been suggested to be top N in recommendation list. The system has been evaluated using LDOS-CoMoDa dataset. In conclusion, it can offer promising approach in recommendation system field if taking the correlations among content and context of items in consideration.