Damir, Be aware the original POS and chunker models in cTAKES were built using text tokenized the way cTAKES tokenizes text. That might not work as well on text tokenized using a different method.
Tim, is that still true of the current POS models in Apache cTAKES? -- James -----Original Message----- From: Damir Olejar [mailto:[email protected]] Sent: Tuesday, May 19, 2015 10:03 AM To: [email protected] Subject: Re: OpenNLP VS UIMA, general question. Yes, that is exactly what I was asking! Thank you, and sorry for not being more thorough with the questions, as I am new to cTakes. Damir On Tue, May 19, 2015 at 10:51 AM, Miller, Timothy < [email protected]> wrote: > I'm not totally sure I understand your question. But if you are asking > if it's possible to use the clinical-trained OpenNLP models released > with cTAKES without using UIMA, yes, it should be possible. Some of the > cTAKES modules simply wrap OpenNLP APIs, and convert the > POS-tagged/Chunked/Parsed output into the UIMA typesystem. > > Or are you asking if you can write a UIMA annotator that consumes > OpenNLP annotations rather than cTAKES' UIMA-based annotations? That is > possible too, though would definitely add complexity to a UIMA pipeline. > > Tim > > On 05/19/2015 10:08 AM, Damir Olejar wrote: > > To whom it may concern, > > > > I would like to ask whether it is possible to have a code written for > > OpenNLP and then, if necessary, integrate it with UIMA. Furthermore, is > it > > possible to go from UIMA to OpenNLP ? For example, I am interested in a > > medical analysis with cTakes, but I cannot find a way how to do it using > > only the OpenNLP. > > > > The reason why I want to rely on OpenNLP as much as possible, is simply > due > > to a complexity of applications I am developing, and UIMA would simply > > complicate everything without a necessity. > > > > Thank you kindly for your answers! > > > > Damir Olejar > > > >
