This will add all of the current unique recommender code into main and also handles some updates to the general kd-tree search algorithm. This will also fix the following three issues:
[x] #5: The title of the issue, sends the top ("best" is currently defined as the closest to the user's centroid (mean of all the documents they have seen so far)) three (can easily be changed to k choices) documents which a given user has not seen.
[x] #6: This fixes small issues with the previous version that wasn't quite as accurate. Although this newer algorithm is somewhat slower (looks at far more of the subtree of the kd-tree), the results seem to be more accurate (more testing needed).
[x] #7: This doesn't entirely solve the issue since this issue isn't really a full issue. The reason for question marks showing up in certain places (especially in the terminal console) is due to issues with printing non-ASCII characters to the console. The smaller answer to this question is to use the function find_and_replace to replace certain HTML encoded values like with a normal space .
Along with the previous fixes, this also has some small fixes to bugs found during testing. Moving forward, more testing will be done on these features to ensure no major issues further down the road.
This will add all of the current unique recommender code into main and also handles some updates to the general kd-tree search algorithm. This will also fix the following three issues:
k
choices) documents which a given user has not seen.find_and_replace
to replace certain HTML encoded values like
with a normal spaceAlong with the previous fixes, this also has some small fixes to bugs found during testing. Moving forward, more testing will be done on these features to ensure no major issues further down the road.