The normal way data is managed is by putting it into the CAS. This allows the flexibility UIMA provides to run in various modes - all in memory in one JVM, or having parts of the pipeline spread out among remote servers.
If you are willing to restrict this flexibility to running in only one JVM on one machine, then you can use external resources to hold arbitrary Java objects and have them made available to the pipeline's components. So, to answer your question "Is this possible?" the answer is yes (using external resources), but with the above restriction. Alternatively, you can put the data into the CAS itself. -Marshall On 1/4/2012 8:41 AM, [email protected] wrote: > Hi, > > I'd like to instanciate an object, assemble the pipeline, run it, access > the object from in an annotator running in the pipeline, and process the > data in that object after the pipeline has finished processing. Is this > possible? Or asked differently: Are files the only way to put data into > a pipeline and get results out? Using external resources, only class > names are given to an annotator as parameter and the annotator > instanciates the object, which is lost after the annotator finished its > job. > > Regards, > > Armin >
