You can implement a custom CandidateItemsStrategy. This class determines which items are initially selected for the recommendation process.
--sebastian On 04.10.2011 16:56, Octavian Covalschi wrote: > Hi there. > > I'm working on a small project that is using taste for user based > recommendations. We were able to integrate MongoDb with it using > MongoDBDataModel instead of file feeds. Results are pretty good in our > opinion, however we'd like to go further and add some kind of filters, to > have even more relevant results. For example each item may have some > attributes and we'd like to use them part of results... I don't want to > apply filters after getting results from taste since we may end up by > excluding all of them, which will make us querying taste again, if that > makes sense... > > So, is there anything out of the box that could help us? Or we'll have to > tweak out datamodel in order to accomplish that? > > Any suggestions are welcome. Thank you in advance. >
