Hi Pascal, Following up on the reply to Eric, you can create an aggregate containing CAS multiplier components that do exactly what you want, but the base UIMA Analysis Engine framework is single threaded: no matter how complex an aggregate AE, only one CAS will be processed by one component at a time. The CPM creates a multithreaded environment that scales vertically within a node pretty well, but does not scale horizontally across nodes. The UIMA JMS design improves on both vertical and horizontal scaling.
Regards, Eddie On 9/25/07, Pascal Coupet <[EMAIL PROTECTED]> wrote: > > Hi Eddie, > > Another example: > > You want to crawl a directory that contains XML Medline files. Each of > these files contains several Medline documents, each of them to be put into > a CAS. We would like to be able to have a clear separation between the > crawling of the directory and the conversion of the Medline XML format. > Doing that it will be easy to get Medline documents from ftp by just > changing the source connection part or to treat a different type of files by > just changing the document conversion part. > > Best, > > Pascal > >
