Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

rewriting rule 3: "vending_machine" #58

Open
vcvpaiva opened this issue Feb 15, 2017 · 6 comments
Open

rewriting rule 3: "vending_machine" #58

vcvpaiva opened this issue Feb 15, 2017 · 6 comments

Comments

@vcvpaiva
Copy link
Member

As the careful analysis of 380 SICK sentences shows (#57) we need to reconstruct mwes before trying to map them to SUMO. Easy examples are "baby panda" ---> not a HumanBaby + Panda, "cartoon airplane" is not an airplane.

@fcbr, @arademaker any suggestions here, once you are given the nn or mwe label?

@vcvpaiva
Copy link
Member Author

Similarly we need new additions for "particle verbs" and prepositions (@arademaker any suggestions from Francis Bond?)

@vcvpaiva vcvpaiva changed the title post-processing CoNLL: fixing multiword expressions post-processing CoNLL2: fixing multiword expressions Feb 20, 2017
@arademaker
Copy link
Member

@vcvpaiva rules for post-processing the dependencies nn -> nn can suggest mwe candidates for example? But as @fcbr said, we need some KB knowledge too. this is a good topic to discuss not here.

@arademaker
Copy link
Member

I didn't understand the second comment.

@vcvpaiva
Copy link
Member Author

@arademaker the dependencies called noun-noun in SD will produce candidates for mwes.
some are crystalized and might be found in a resource such as PWN (or an extension of PWN that paid more attention to compounds like ARK). e.g. paddling pool, vending machine, soccer player.

others are wrong, like "man eating" in "There is no man eating", where "eating" is not a noun.

others are not crystallized (like forest path) and are difficult to decide how to deal with them (use both concepts or just the head?) . finally, others still, like "tiger cub" I cannot even tell if it's crystalized or not.

@vcvpaiva
Copy link
Member Author

vcvpaiva commented Feb 21, 2017

about my second comment it has two parts:
1.you mention that Francis might be involved with a LKB (lexical knowledge base) of prepositions?
if we don't go the whole hog looking for such, we should at least use the ones in the SDs manual
http://nlp.stanford.edu/software/dependencies_manual.pdf as a post-processing step.
2. we need to decide how to deal with particle verbs. maybe more can be added to PWN, but if not, some need to not produce concepts. the example I gave "the toddler is standing up". "up" is just a particle that determines the sense of standing, but if in "the index is up", "up" should map to "Increasing".

issue #65, issue #36

@vcvpaiva vcvpaiva changed the title post-processing CoNLL2: fixing multiword expressions rewriting rule #3: creating multiword expressions for "paddling pool" and "vending machine" Mar 15, 2017
@vcvpaiva vcvpaiva changed the title rewriting rule #3: creating multiword expressions for "paddling pool" and "vending machine" rewriting rule #3: "vending_machine" and "swimming_pool" Mar 15, 2017
@vcvpaiva
Copy link
Member Author

As commented above there several kinds of multiword expressions that we need to have.
this issue has now been restricted to mwes that exist in PWN, like vending_machine and swimming_pool. unfortunately paddling_pool does not seem to exist in PWN

@vcvpaiva vcvpaiva changed the title rewriting rule #3: "vending_machine" and "swimming_pool" rewriting rule #3: "vending_machine" Mar 15, 2017
@vcvpaiva vcvpaiva changed the title rewriting rule #3: "vending_machine" rewriting rule 3: "vending_machine" Mar 16, 2017
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants