NLP Parse Tree
[sorry for my English]
Is it possible to get NLP parse tree as a part of the evidenceList in the answer or by optional request?
Say, for the evidence phrase:
"Watson can read and understand natural language, important in analyzing unstructured data that make up as much as 80 percent of data today."
The parse tree is (produced by nlp.stanford.edu:8080/corenlp/process):
(ROOT (S (NP (NNP Watson)) (VP (MD can) (VP (VB read) (CC and) (VB understand) (NP (NP (JJ natural) (NN language)) (, ,) (ADJP (JJ important) (PP (IN in) (S (VP (VBG analyzing) (NP (NP (JJ unstructured) (NNS data)) (SBAR (WHNP (WDT that)) (S (VP (VBP make) (PRT (RP up)) (ADVP (RB as) (RB much)) (PP (IN as) (NP (NP (CD 80) (NN percent)) (PP (IN of) (NP (NNS data))))) (NP-TMP (NN today))))))))))))) (. .)))
As I understand, Watson produces kind of that tree too as result of the internal workflow. We need it for the further processing on our side. In other words we would like to have some parts of our corpus in the structured version too (as the parse tree).
Thanks, Slava V.
Hi Slava,
Will is right - currently you don't get the tree or the NLP as a separate service offering. You work with the JSON response as part of the QA API, that produces metadata other than the answer, evidence and confidence levels.
What specifically are you trying to do downstream with the tree?
This is not a critical problem. Don't worry.
Thanks, Slava V.
Hi Slava, good to hear from you again. As part of the API, you'll get the ranked answers, confidence levels, and evidence back as your response. I don't believe you'll be able to extract what you're asking for. Let me check on this one to validate.
:) me too.
Thanks Will, SridharS. We can live without that tree.
Slava V.