Exercise 8, Grammar Development

Complementation and Control

Control: COMPs and XCOMPs

Integrate COMPs and XCOMPs into your grammar. The following sentences should work:

You will need to expand your grammar rules along the lines shown below. Your VP rule will, of course, be more complex than that shown here since you need to be able to do simple sentences as well (all the types that you can already do).

Remember to do Regression Testing before, after and while you are expanding your grammar! Run the testsuite to make sure that you can still parse everything you want to parse in the right way.

S --> NP VP.

VP --> V  (NP) ({VPinf | CP}).

VPinf --> PARTinf VP.

CP --> C S. 

Hint: below is the tree commonly assumed for control constructions.

              S 
	    /   \
          /       \
        NP           VP 
        |          / \  \ 
      John       V    \   \
	    persuaded   NP  \
	    promised   / \    \
	              D   N    \
                     the cat  VPinf 
                              /  \
                         PARTinf  VP 
                           to    /  \
	                        V    NP
                               eat   |
                                    yoghurt

Note that the c-structure rules have not been annotated with any functional information as yet. You will need to do this. We went through how to implement these in class, so look back, think carefully about what you are doing and make sure you understand how the system is working.

You will also need to add control equations to the lexical entries for persuade, promise, want in the manner that is appropriate. Verbs selecting for a COMP don't contain a control equation (of course).

For example:

said   V * { (^ PRED) ='say<(^ SUBJ)(^ COMP)>'
	       |(^ PRED) = 'say<(^ SUBJ)(^ OBJ)>'}

convince  V * (^ PRED) = 'convince<(^ SUBJ)(^ OBJ)(^ XCOMP)>'
	      (^ OBJ) = (^ XCOMP SUBJ)

try 	  V * (^ PRED) ='try<(^ SUBJ)(^ XCOMP)>'
	      (^ SUBJ) = (^ XCOMP SUBJ)

Recursive Embedding

Make sure that your grammar allows for recursive embedding (it should already). The following sentences should work:

  1. Gabrielle persuaded Joxer to want to play the lyre.
  2. Xena wanted Gabrielle to believe that Joxer promised to play the lyre.
  3. Gabrielle thought that Xena wanted Joxer to think that the horse wanted to eat hay.

Long Distance Dependencies

Long Distance Dependencies

Expand your grammar to deal with long-distance dependencies. (you can read up on them some more in Chapter 14 the Dalrymple textbook).

Introduce an extra Stop (stands for S topic) (Stop --> NP (COMMA) S) and annotate the topicalized constituent as shown below.

Functional Uncertainty Annotation

Add the functional uncertainty equation below to the NP preceding the S.

(^ TOPIC) = ! 
(^ {XCOMP|COMP}* {OBJ|OBJ2}) = !

You should be able to parse (at least) the following sentences.

  1. Ram loves hoppers.
  2. Hoppers, Ram loves.
  3. Hoppers, Ram wants Tara to like.
  4. Hoppers, Ram persuaded Tara to want to eat.
  5. Hoppers, Ram wanted Tara to think that Ravi persuaded Amra to eat.
  6. Hoppers, Ram thought that Tara wanted Amra to think that Ravi wants to eat.

Relevant Reading Material

Dalrymple (2001), Chapter 12, pp. 313-330.

Cookbook, Sections 2.2, 2.3, 3.3-3.5