I know that these tasks themselves are important for theoretical insights.
Show this thread
Naive question: basic NLP tasks such as POS tagging and SRL are useful for applied tasks such as MT and QA? If the input information is just reduced by the basic tasks, it is not obvious that their output is useful for applied tasks. Indeed, it may improve interpretability.
-
-
-
I want to note that I don't like using "useful" in any academic context. This is just like a hypothetical question...
Show this thread End of conversation
New conversation -
-
-
OKay. This question doesn't make sense if we can use POS tagger etc. as external resources for higher tasks.
Show this threadThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Naive answer: maybe you can consider they are adding value by disambiguating things and allowing for generalization from less data in higher level processing? Knowing something is a noun explicitly (assuming it's correct) could help, esp. when you are dealing with a rare word.
-
If the input information used for predicting POS tags (not as an external additional resource in my assumption) is the same as the input information used for higher-level tasks, you might not need the information of POS tags itself. I'm not very sure though...
-
I don't think the information will be the same. Here is how I see it: POS tagger works because it has 'knowledge' extracted from the data used to train it. Using a POS tagger means adding (a small potion of) the 'knowledge' into plain text, at least when it's successfully done.
-
What do you think if we can use the same data (used for training POS tagger) even for higher processing tasks? (In the first place I assumed this, but now I realized that it doesn't make sense...) Anyway, I agree with you that the POS tagger has knowledge. Thanks!
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.