| 摘要 |
The invention involves clusters or hubs each comprising multiple works for which human beings might express taste-based preferences. The items are grouped in clusters in such a way that the works most in accordance with the tastes of any particular individual person will tend to be in a small number of these clusters out of the overall collection. In this way, clusters can be used to help the person find items that he is not previously familiar with but that he will probably like. The clustering of works is optimized by human effort, software, or both. By way of example, a methodology for doing this using the principle of information transfer as described in the theory of Shannon entropy is explained. When human effort is used to perform the optimization, facilities are provided for using such principles to determine whether a human-suggested change actually improves the clustering. Facilities are provided whereby the optimization work may be distributed across multiple machines. Facilities are provided whereby artists may introduce new works to the system and quickly make them known to the people who are likely to enjoy them. Facilities are provided whereby users can easily receive recommendations for works they are likely to enjoy.
|
