4

I'm interested if there is a text or set of texts where each word is correctly POS tagged. I know there are algorithms that can associate POS tags to the words, but there are always many of incorrectly tagged words, which are ambiguous and context sensitive. These errors could be only handled at the time the text is created by some clever editor and the author of the text would clarify the POS tags of words that could be ambiguous in particular context. So, the result would be the text where all POS tags are correct.

Could you recommend some research material dealing with creation (possibly computer AI assistance) of unambiguous texts by some clever editor which could be further processed on the semantic level?

Is there such a corpus or set of texts available on the Internet, which are created by this tool?

Improve this question
5
  • you have asked 2 questions whose answers won't be related to each other. It would be better if you either asked 2 separate questions on this SE, or add context that relates the two questions. – prash♦ Oct 30 '12 at 8:36
  • The main question is for tool that helps the writer of text to show ambiguous constructs and if there is some repository of unambiguous POS tagged texts created by this tool, it would be even better. – xralf Oct 30 '12 at 8:45
  • 1
    The Brown Corpus is one example. I'm sure there are more if you google for it. – Mitch Oct 30 '12 at 15:25
  • 1
    Q2: Depending on what kind of ambiguity you're thinking of, my own limited understanding is that ambiguity at all levels is so pervasive that it is impossible to come up with a completely unambiguous natural language text. I believe that in any manual tagging task there is disagreement among taggers, which can then potentially be largely (though not completely) resolved through discussion. Even if you had the author himself trained and tagging the text, the tagset itself can be just slightly too fine/coarse-grained, introducing more "ambiguity". – lapropriu Oct 30 '12 at 22:28
  • 1
    Asking for a manual POS tagger is I think a good question. Such tools are necessary if you want to train an automatic POS tagger - it has to have something to gauge its success. But it's false to assume that a manually POS-tagged text would be ambiguous. Languages are ambiguous, POS are not set in stone - they're human created and fallible. There can be more than one good set of POS for a language and more than one way to apply them to the same text. – hippietrail Oct 31 '12 at 12:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.