NLP POS Annotation Tool with Penn Treebank tags [closed]

Closed. This question does not meet Stack Overflow guidelines. It is not currently accepting answers.

---

This question does not appear to be about a specific programming problem, a software algorithm, or software tools primarily used by programmers. If you believe the question would be on-topic on another Stack Exchange site, you can leave a comment to explain where the question may be able to be answered.

Closed 9 years ago.

开发者_运维问答 Improve this question

I've got a corpus that I want to annotate the parts of speech (verbs, nouns, adjectives, etc.) I'm looking for a nice tool that I can use to do that however I have a requirement that I want it to tag the corpus using the same tags that Penn Treebank does. The reason is that I want to use Stanford NLP to do the pos identification.

Any help is appreciated,

mj

---

You might want to look into brat or webanno. These are both span-based annotators rather than token-based annotators but you should be able to double-click words and then pick the tags from a list if they're configured correctly.

brat has a dependency and tagging example configuration which would probably be a good starting point, though you may want to borrow parts from the Stanford CoreNLP configuration's visual.conf which includes some coloring for Penn Treebank tags. I'm less familiar with webanno's configuration but since it's based on brat, one can presumably customize it in the same way.

---

You have a number of options; here's my thoughts in order of easiest/least complex to most complex:

1) Any of the modules on CPAN, particularly Lingua::EN::Tagger. Not sure how accurate it is, but it's very very easy to implement, but you have to like Perl.

2) Almost as easy NLTK (a python natural language toolkit). Installing the whole package takes a while, but writing the code is easy cheezy. NLTK has pretty strong documentation and examples; here's pos tagging: http://nltk.googlecode.com/svn/trunk/doc/book/ch05.html

3) Personally I tend to use the stanford parser a lot and there's a nice pos tagger built in. I have an example of programmatic interaction with the stanford system on my webpage (search the page for 'Simple version using built in tokenization'): https://sites.google.com/site/nicoflacco/ You'll have to customize the code a bit to only do the tokenization/tagging and not parse, but that's not too hard.

4) Lingpipe is a bit heavier duty. I believe they incorporate the stanford parser, but I could be wrong.

From what you are saying (3) might not be bad if you want to use the stanford parser and not just the pos tagger.