identitfy negated contexts in text

Hello everybody,

after my first experiences with KNIME, I am desperately looking for a solution to identitfy negated contexts in text. The POS tagger tags the individual words, but without dependencies.


Example (Stanford POS):
I don't require this knowledge.
I(PRP) do(VBT)n't(RB) require(VB) this(DT) knowledge(NN).

Is there a possibility to create a dependency tree to identify negated contexts in text:
Example:
(ROOT
 (S
  (NP (PRP I))
    (VP (VBP do) (RB n't)
     (VP (VB require)  (NP
      (DT this) (NN knowledge))))
(. .)))


--> neg(require-4, n't-3)

I haven‘t found anything to this topic in the forum and I am grateful for any idea.
Best regards
Matthias

Hi Matthias,

That is not possible with the nodes currently included in KNIME. When converting documents to a bag of words the order and also the sentence boundaries get lost (in the bow not in the docs themselves). The only way I see is that you implement a custom node that operates on documents and extracts relations triplets.

The Stanford NLP lib provides functionalities for triplet and relation extraction as far as I know.

Cheers,

Roland