This problem is rather major. Currently there are some issues that arise when testing the searching algorithm:
Somehow there still seem to be duplicates (though rare). Some nonce check fails when trying to find if a user has already seen that document. Testing and research into the algorithm is required.
When a new user (a person who has seen one document) joins the system, the results all start at the same position and only slightly shift based on viewed content over time. The algorithm may need further generalization to ensure enhanced results.
Also see #6 for further details of KD-tree search adjustments.
This problem is rather major. Currently there are some issues that arise when testing the searching algorithm:
Also see #6 for further details of KD-tree search adjustments.