Hi Sumit, Thanks for your suggestion, it seems like the proper way to go for my use case. However, I'm not too familiar with the UIMA internals, so could you point me to where or how I can set the dropCasOnException option? Thanks! Carsten
Am 07.11.2014 um 10:19 schrieb Sumit Madan: > Hi Carsten, > > I had this experience too that a flow controller is not easy to build. > But may be you can use a workaroud. You can put a new AE in-between > (BinaryCasReader and Segmenter). This AE would throw an exception when a > (J)Cas doesn't fit your rules. With the UIMA options dropCasOnException > and ActionOnMaxError, UIMA can drop those (J)Cases and go further with > the wanted ones. > > Regards > Sumit > > On 07/11/14 09:04, armin.weg...@bka.bund.de wrote: >> Hi Carsten, >> >> I've never used it, but according to the documentation you can do this >> with a flow controller. The bad thing is, Richard told me a while ago >> that it is not so easy to build your own flow controller. >> >> Cheers, >> Armin >> >> -----Ursprüngliche Nachricht----- >> Von: Carsten Schnober [mailto:schno...@ukp.informatik.tu-darmstadt.de] >> Gesendet: Donnerstag, 6. November 2014 14:55 >> An: user@uima.apache.org >> Betreff: Filter Cas from UIMA fit pipeline >> >> Hi, >> I wonder whether there is a recommended way to remove certain (J)Cas' >> (i.e. documents) from a pipeline after reading. >> The scenario in my case is that I use a standard reader >> (BinaryCasReader) which returns many documents. I only want a subset of >> these documents to be processed by the following pipeline (comprising a >> segmenter, a writer and some other engines), subject to a certain value >> in a custom annotation. >> >> The initial intuition would be to use/implement a reader that only >> selects those documents that fulfil the given condition. In my case that >> would mean, however, that I'd need to implement a new Reader extending >> the BinaryCasReader by the described functionality. From a high-level >> view at least, this seems much more complicated than just removing >> documents from the pipeline. >> Can I avoid that effort somehow without breaking conventions? >> >> Thanks! >> Carsten >> > > -- Carsten Schnober, M.Sc. Doctoral Researcher Ubiquitous Knowledge Processing (UKP Lab) FB 20 Computer Science Department Technische Universität Darmstadt Hochschulstr. 10, D-64289 Darmstadt, Germany phone (0)6151 16-6227, room S2/02/B111 www.ukp.tu-darmstadt.de