I think that intuition and observation are crucial factors that strongly affect the proposed methods/approaches solving the problems in NLP research. I read quite a lot of NLP papers and recognized this. Due to the ambiguity in natural language, some of NLP problems will be heuristically solved based on some intuition and observation from what humans are able to do naturally.
Anyway, just my opinion. It is quite subjective. Please contradict me if any!
--
Cheers,
Vu
Anyway, just my opinion. It is quite subjective. Please contradict me if any!
--
Cheers,
Vu
Not too wrong, I must admit.
ReplyDeleteIn my limited exposure to research in language technologies, I felt the same.
It turns out to be so very true in summarization, for example. A simple position based baseline algorithm performs better than most theoretically and methodologically strong frameworks.
And in evaluation, ROUGE-2 correlates so well to human evaluation, I often wonder how complex the problem is: or is it that we don't understand the problem that well, yet.
--
Rahul