Method and system for recommending geo-tagged items

Abstract

A recommender system for recommending items to a user based on geo-Tagged information related to him, in which items associated with a GeoTag are stored in a database. Feedback regarding the various items is obtained from the user and the provided rating of items is propagated to closely located items based on their associated GeoTags. A user-to-user similarity matrix is calculated and a predicted score is assigned for each user and item, using a recommendation server. All the items in the catalog of items are sorted according to their predicted scores as calculated for the user, and all items that have been already rated by the user are filtered out. Then, items from the catalog of items are presented to the user, according to their scores.

Description

CROSS-REFERENCE TO RELATED APPLICATION[0001]This application claims priority from Israeli Patent Application No. 212502, filed Apr. 26, 2011, which is incorporated by reference herein in its entirety.FIELD OF THE INVENTION[0002]The present invention relates to the field of recommender systems. Specifically, this invention relates to a recommender system which is designed to recommend objects to a user based on Geo-Tagged information related to him.BACKGROUND OF THE INVENTION[0003]With the escalating amount of data available online, recommender systems became very popular, especially on web sites. As known in the art, recommender systems are systems that recommend items to users. Such systems have various applications such as helping users find web pages that interest them, recommending products to customers in e-commerce websites, recommending TV programs to users of interactive TV and displaying personalized advertisements. There are many types of recommender systems ranging from m...

AI Technical Summary

Problems solved by technology

Recommender Systems are supposed to be used by people that do not have sufficient personal experience or competence to evaluate the, potentially overwhelming, number of alternatives offered in a web site.

In many cases such solution prevents the user from receiving recommendations regarding items from a far surrounding which may be of high interest to the user.

Additionally, some traditional methods require to infer the preferences of the user from the user's contextual information which in many cases is not related to the preferences of the user and in some cases may even deceive, for example if a user writes “I would prefer any thing other then Pizza”, a context-aware recommender system may infer that the user prefer Pizza and recommends him pizzerias in his geographic proximity.

However they might neglect obvious geo-relations.

However it will not be able to recommend news coming from the city “Darmstadt” even if it is only 20 km away.

However, these kinds of solutions are usually user based and are only capable of recommending articles located closely to previously clicked news and cannot be used to recommend entirely new locations.

However, it is not practical to simply refer to the geo-location as “items” like in any other CF application, because in fine-grained application there will be tremendous number of locations or even a much higher number of geo-location information that might be pointing to the same location.

Images