Yes, most of the DKPro Core components rely on token/sentence annotations. You can find a list of the types and which components create/consume them here https://dkpro.github.io/dkpro-core/documentation/ under the section "DKPro Core 1.8.0-SNAPSHOT" -> "Typesystem reference".
Best, -- Richard > On 16.02.2016, at 13:13, Nils Reiter <[email protected]> wrote: > > Hi Richard, > > thanks for your reply and don’t worry, I am planning on using DKpro > components :) > > So if I get you correctly, all DKpro components rely on token/sentence > annotations and ignore the rest, right? > > Best regards, > Nils > >> On 16 Feb 2016, at 12:18, Richard Eckart de Castilho <[email protected]> wrote: >> >> Ok, sorry, the answer below would assume you are using DKPro Core components >> ;) >> >> Sorry Nils, I didn't notice you were posting to the Apache UIMA list. >> >> So for UIMA in general, I am not aware of a solution other that what you >> describe. So it would depend on the components / component collection that >> you are using. >> >> Cheers, >> >> -- Richard >> >>> On 16.02.2016, at 12:17, Richard Eckart de Castilho <[email protected]> wrote: >>> >>> The easiest would be to remove the token/sentence annotations of those >>> parts of the text that you do not care about. >>> Or alternatively - if you have annotations that specifically mark the text >>> sections, then configure the segmenter component to create sentences/tokens >>> only within the boundaries of these annotations using PARAM_ZONE_TYPES and >>> PARAM_STRICT_ZONING. >>> >>> Cheers, >>> >>> -- Richard >>> >>>> On 16.02.2016, at 12:02, Nils Reiter <[email protected]> wrote: >>>> >>>> Hi, >>>> >>>> is there a way to run an analysis engine on only a part of the CAS? >>>> >>>> I have UIMA annotations over all the substrings that I want to process. >>>> The only way I could think of is creating new views or CASs for each >>>> string, but that would result in > 100 views. Is there a more >>>> straightforward way? >>>> >>>> Background: >>>> Only part of the CAS contains natural language, other parts are lists, >>>> names and headers. I would like to POS-tag the text, but not the rest. >>>> >>>> Thanks in advance for any pointers or suggestions, >>>> Nils >>> >> >
