I can really recommend the DKPro Lab that Richard suggested! -- ------------------------------------------------------------------- Oliver Ferschke, M.A. Doctoral Researcher Ubiquitous Knowledge Processing Lab (UKP-TU DA) FB 20 Computer Science Department Technische Universität Darmstadt Hochschulstr. 10, D-64289 Darmstadt, Germany phone [+49] (0)6151 16-6227, fax -5455, room S2/02/B111 [email protected] www.ukp.tu-darmstadt.de Web Research at TU Darmstadt (WeRC) www.werc.tu-darmstadt.de -------------------------------------------------------------------
________________________________________ Von: Richard Eckart de Castilho [[email protected]] Gesendet: Montag, 17. Juni 2013 21:36 An: [email protected] Betreff: Re: Processing a Text Collection more than once? Hi Susanne, there are two options in UIMA: 1) you write your own reader which repeatedly outputs the same data 2) you write a flow controller which saves all data produced by the reader and re-runs all components again on it However, I'd recommend we check offline if/how DKPro Lab could fit in with your scenario. It may even just be the easiest to run your pipeline in a loop using uimaFIT and XMI/binary CAS serialization to feed in the output of one run into the next one. Cheers, -- Richard Am 13.06.2013 um 17:32 schrieb Susanne Neumann <[email protected]>: > Hi, > > is there a (good) way in UIMA to process the whole text collection more than > once? The process method processes each document once for the whole > collection. But I need to iterate several times over the whole collection. > > The background is, that I want to implement a bootstrapping annotator using > UIMA. One of the main characteristics of bootstrapping is, that the corpus is > processed several times, collecting new rules, terms and evidence each time, > based on the results of the previous turns. I planned to write a > bootstrapping AE, but I can't figure out how to iteratively process the > collection. > > I am looking for any hints or tips about how to implement this with UIMA. > > Thanks, > Susanne
