Abstract
Recommendation system based on the actions of a group of users, and not requiring prior metadata, is provided. The recommendation is not limited to from user-to-item (e.g. media recommendation) type; e.g. from user-to-user (e.g. social networking), from item-to-item (e.g. "see also" in internet shopping) and item-to-user (e.g. advertizing) recommendations are supported. The solution utilizes a set of identifiable tokens, associated with each entity, an entity being either a user or an item. Whenever there is an interaction between entities, either existing token sets are updated or new token sets are created such that after an interaction each interacting entity is associated with a token set which resembles more than before the token set(s) associated with the other interacting entity. Recommendation can be produced by searching best matching token sets on the user-to-item, user-to-user, from item-to-item or item-to-user basis.
Patent family as of 21.10.2021
FI126426 B 20161130 FI20120006249 20121128
FI20126249 A 20140224 FI20120006249 20121128
US2014059213 AA 20140227 US20120687237 20121128
US9882790 BB 20180130 US20120687237 20121128
WO14029904 A1 20140227 WO2013FI50687 20130620
Link to current patent family on right
Original language | English |
---|---|
Patent number | US2014059213 |
IPC | H04L 12/ 26 A I |
Priority date | 28/11/12 |
Publication status | Published - 27 Feb 2014 |
MoE publication type | H1 Granted patent |