DIGITAL CONTENT PROVISION

摘要

A collaborative recommender system which is based on an application of Formal Concept Analysis is disclosed. A problem with collaborative recommender systems based on Formal Concept Analysis is that the size of the concept lattices generated can be very large. Large lattices both take a great deal of time to generate, and tend to include extraneous concepts. In the collaborative recommender systems disclosed herein the size of the concept lattices used are only of a scale required to capture the tastes of an individual user. This reduces the computational resource required by the recommender system markedly. Further improvements are obtained in some embodiments by selecting only relevant concepts from the concept lattice before building a user profile from the selected concepts. Yet further improvements are obtained by the introduction of a relevance measure which depends upon the difference between the size of the extent of the concept whose relevance is being measured and the size of the largest lower neighbour of the concept whose relevance is being measured. This relevance measure can be computed more easily than prior relevance measures.