An important and challenging task in any keyword-based search system in text documents or relational databases is the capability of the system to find additional results besides the actual search results and present them to the users as recommendations. This function allows the records that might be of interest to the user to be discovered and essentially enhances the user's browsing experience. Most recommender systems such as Amazon and IMDB rely heavily on the users' ratings, previously learned patterns from the users and their selected items to achieve this goal. In this paper we present a system called Tuple Recommender which first searches a relational database for a given keyword query and then makes the search recommendations based on the similarity of the tuples with respect to the tables' attributes in which the search terms are found, without relying on the previously learned patterns or users' ratings.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below