On 20 March 2014 10:15, Dileepa Jayakody <[email protected]> wrote:
> > In the ReputationBox, another main component is the reputation-analysis > component. > [snip] > The best opensource option to implement a recommendation engine I have seen > so far is Apache Mahout. > Had a quick look over the Mahout website [1] and read a blog or two [2]; does look like a good fit. > > Any suggestions, pointers on how to integrate a Mahout machine-learning > process to do the email analysis within the application will be greatly > appreciated. > > Seems that there are two phases to this... the initial bulk extract and learning exercise, and then on-the-fly scoring. I suppose that the initial bulk extract and learning could be done as a background task in Isis... but I'd probably spike it outside of Isis in the first place. The on-the-fly scoring looks easy enough; a synchronous call from the RB gmail client to the RB server hosted by Isis and its REST API. I think overall this might be more an a Mahout- than an Isis project; but it'd be nice to combine all of these (plus Shiro for the auth stuff, of course). Other opinions on this topic welcome... HTH Dan > Thanks, > Dileepa > > > [1] https://mahout.apache.org/ [2] http://emmaespina.wordpress.com/2011/04/26/ham-spam-and-elephants-or-how-to-build-a-spam-filter-server-with-mahout/
