The Stanford NLP group maintains an online demo that uses the Stanford Parser to deliver phrase-structure trees and Stanford dependency graphs of free-form text that users enter:
Your task: at the parser demo,
Negation is extremely important semantically and extremely complex in its syntactic expression. The following is a sample of negating words and constructions in English:
not, never, nothing, no as in No, it's raining, no as in no student, n't, nowhere, none, few, seldom, rarely, have yet to, deny, ...
Your task: Using the parser demo or the Stanford dependencies manual, identify three distinct ways in which negation is expressed in collapsed Stanford dependency structures. Feel free to extend the above list of negations as you see fit (the more the merrier!).
A bit of background: Syrett and Lidz (2010)* report on a corpus study of adverb–adjective combinations in English, seeking to find patterns that can be traced to interpretive restrictions and preferences. For example, half full sounds normal and is easy to interpret, whereas half tall is odd and hard to interpret. This pattern presumably traces to the sort of scales along which we measure full and tall. (You might mutter to yourself different combinations of completely, somewhat, and mostly with full, smart, and damp to get a feel for what the patterns are like.)
Your task: propose a matrix design (as discussed in class on Jan 17) that (i) makes use of Stanford dependency structures (regular or collapsed) and (ii) could be used to provide a data-rich picture of what the patterns of adverb–adjective modification are like. Describe the design by answering at least the following questions:
In addition, articulate one specific research question that your design would help address. (Try to be creative here; this task could stimulate new research!)