The traditional NLP/UIMA pipeline model fits very well with storm's spouts/bolts model. I built out something for myself which has uses isolated analysis engines for different tasks and built some spouts out as adapters. The basics of it would be doing a process on each analysis engine, and then using one of the cas serialization utilities to pass results along each spout. You could load the text you want to analyze via any number of means (JMS, REST,...) into a spout which will then forward it to the rest of the topology. I've found it to be great for my use case.
Hope that helps! On Sat, Dec 21, 2013 at 10:20 PM, Vinod <[email protected]> wrote: > Hello, > I recently came across Storm (twitter storm) and was wondering if it is > possible ( and even reasonable ) to deploy an UIMA pipeline in Storm. > > Any comments and/resources would be greatly appreciated. > > Thanks, > Vinod. > >
