Thanks for sharing your thoughts guys. I think it would be better for me to keep the two layers separate. The UIMA pipeline can be used to extract useful features. Another layer can then use those features to implement and generate deep learning models (via mahout/mapR jobs)
Cheers Som On Fri, Feb 15, 2013 at 6:48 AM, Brian Dolan <[email protected]> wrote: > We tackled this same issue. Ultimately, since a UIMA process is usually > concerned with a single document, it made more structural sense to wrap the > UIMA task within a Mapper. That keeps the entire process within the > functional programming paradigm. We also were concerned with how fragile > the UIMA configuration can be and it was easier to control when embedded > within a Mapper. Similarly with Mahout, though we separated the two jobs. > > > On Feb 15, 2013, at 2:37 AM, Julien Nioche <[email protected]> > wrote: > > Hi > > I suppose you could expose MapReduce jobs as UIMA components but it would > certainly be easier to do the other way round and use e.g. Behemoth [1] to > run the UIMA PEARs on MapReduce. > > HTH > > Julien > > [1] https://github.com/DigitalPebble/behemoth > > On 13 February 2013 22:47, Som Satpathy <[email protected]> wrote: > > > Hi all, > > > > I have been toying around with UIMA pipelines for some time now. I was > > wondering if UIMA can support analysis components written as mahout > > map-reduce jobs as part of a UIMA pipeline ? > > > > I would appreciate any help/hints/pointers. > > > > Thanks, > > Som > > > > > > -- > * > *Open Source Solutions for Text Engineering > > http://digitalpebble.blogspot.com/ > http://www.digitalpebble.com > http://twitter.com/digitalpebble > >
