LING0001 -- Homework 5

(Due 10/28/2022)

1. Parsing simple phrases

It's traditional for many LING001 students to have trouble with the basic concepts of syntactic analysis, and translating these concepts to "labelled bracketings" and "tree structures". So this homework assignment is intended to lead you towards better understanding of these issues, through a set of simple examples followed by some slightly less simple problems.

We'll start with some exposition -- if you've already figured this stuff out, you can skip ahead to the actual questions.

If we give the word sequence "University of Pennsylvania" to one of the online constituency-parser demos, it tells us that the structure is

(NP (NP (NNP University)) (PP (IN of) (NP (NNP Pennsylvania))))

Or, in the equivalent graphical format that most people find easier to read:

The formula and the graph are attempting to apply the constituent tags and the parsing principles of the Penn Treebank, and in this case, the details express (among other things) the idea that

If we change all the parentheses to square brackets,

[NP [NP [NNP University]] [PP [IN of] [NP [NNP Pennsylvania]]]]

we can use the online jsSyntaxTree program to produce an equivalent diagram, as shown below.

We could make the bracketed formula a bit more readable by "pretty printing" it:

[NP
  [NP [NNP University]]
    [PP [IN of]
      [NP [NNP Pennsylvania]]]]

There are other systems of constituent-structure labels and formulas and graphs -- try e.g. this one -- but they all will agree that "University of Pennsylvania" is a noun phrase (perhaps using a different label), that "of Pennsylvania" is a prepositional phrase (again maybe under a different label), that "of" is a preposition (under some label), and that "University" and "Pennsylvania" are nouns of some kind.

For the purposes of this course, you don't need to learn any particular system of syntactic terminology. We just want you to learn the basic idea of syntactic analysis, which is that human languages create more complex messages as structured combinations of simpler ones, and that the structure of the combination matters.

Formats and node-label details aside, this first example doesn't tell us a lot about syntactic analysis, since there's basically only one way to put a NOUN-of-NOUN sequence together. But if we add another word before or after, more options open up. Thus

University of Pennsylvania campus

is the campus of the University of Pennsylvania, not a university known as "Pennsylvania campus". And Indiana University of Pennsylvania

is

[NP [NP [NNP Indiana] [NNP University]] [PP [IN of] [NP [NNP Pennsylvania]]]]

not

[NP [NP [NNP Indiana [NP [NP [NNP University]] [PP [IN of] [NP [NNP Pennsylvania]]]]

Many phrases are ambiguous, for humans and even more often for automatic parsers. This is especially true for headlines. Thus the same parser we used before, given the headline "Inflation risks dog economy", analyzes risks as a verb, and dog as a noun, as if the headline meant something like "Inflation threatens (a) depression economy":

[S [NP [NN Inflation]] [VP [VBZ risks] [NP [NN dog] [NN economy]]]]

whereas it should treat risks as a plural noun, and dog as a verb, meaning something like "Inflation problems threaten (the) economy":

[S [NP [NN Inflation] [NNS risks]] [VP [VBP dog] [NP [NN economy]]]]

This change affects both parts of speech (is risks a verb or noun, is dog a noun or verb?) and also constituent structure (does risks link to the right, as a verb governing the noun phrase "dog economy", or to the left, as the head of the noun phrase "inflation risks"?).

2. The Homework Problems

OK, on to the actual problems!

Analyze the ambiguity of five of the following nine (actual) headlines:

  1. Woman abandoned as newborn searches for birth mother
  2. Does Donald Trump support matter?
  3. Hospitals named after sandwiches killed five
  4. Missing woman remains found
  5. Police kill unarmed deaf man using sign language
  6. Dead mouse in protein supplement claimant admits lying
  7. Trump demands dog 'Dreamers' deal
  8. Chinese cooking fat heads for Holland
  9. Violinist injured in JAL crash blossoms

In each case