Reminder of commands for using the system

• To run the parser on an input file with feature unification:

  $ runparser +c in > out

• Browse the parse forest:

  $ xtag.browser

  Then load a parse forest produced before.

• Inspect the derived trees with features:

  $ print_deriv -p -f out | showtrees    

  Inspect the derivation trees:

  $ print_deriv -d out | showtrees    

• Inspect tree schemata with features:

  $ xtag.show english ^alphanx0Vnx1$

• Inspect elementary trees:

  $ xtag.show.word english likes "alphanx0Vnx1\[likes\]"

Basic exercises:

• Have a look at the analyses of the example

  (1) Peter likes Mary .

  Where does the ambiguity come from; explain the second analysis.

• Have a look at the analyses of

  (2) Mary takes a bath .

  Explain the difference between the two analyses with determiner a (not noun a).

• Have a look at the analysis of

  (13) the music should have been being played .

  The auxiliaries all use elementary trees with the tree schema betaVvx. How do the actual elementary trees differ? Which feature controls the ordering of the auxiliaries?

Sentential complements:

• Have a look at the analyses of the example

  (10) Peter expected him to come .

  There is one analysis with expected as an ecm verb; in this case the elementary tree for expected is betaXnx0Vs1.

  Which feature makes sure expected assigns case to the subject of the embedded infinitive? (Look at the elementary trees for expected and for come.)

  Now have a look at example

  (11) Peter expected he to come .

  Why is no ecm analysis available for (11)?

• Have a look at the analysis of the example

  (12) Peter wanted to come .

  Which feature expresses the fact that want is a subject-control verb?

• Compare the analyses of (14) and (15):

  (14) he hopes that Peter wins .

  (15) that Peter wins annoys him .

  Consider only those analyses where that is a complementizer. How does the analysis of sentential subjects differ from the one of sentential complements? (Look at the derivation trees.)

Wh-movement and subject-auxiliary inversion:

• Have a look at the analyses of the example

  (3) whom does Peter like .

  Choose the correct analysis (NP extraction, not S extraction).

  Which feature establishes the link between the moved wh-element and its trace?
  How does the verb assign case to the wh-element?

• Run the parser with

  (4) whom Peter likes .

  Look at the output, you will see that the parser does not find an analysis.

  Then have a look at the analysis of

  (5) Peter, Mary likes .

  In both cases, the elementary tree for likes is the same as in the analysis of (3). Additionally to the usual feature unifications, a final unification of the invlink and wh features of the top S node of the matrix clause take place.

  Explain why (4) cannot be parsed while (5) is grammatical.

• Have a look at the elementary tree of comes in the analysis of

  (6) who comes . 

  Why is it possible here to have no inversion? Is inversion still allowed? (Take a look at the analysis of (7):)

  (7) who does Peter think comes .

Relative clauses:

• Have a look at the analysis of the example

  (8) the man who arrives laughs .

  Now try

  (9) the man who arriving laughs .

  You will see that there is no analysis with an arriving relative clause modifying man.

  Why is this the case? (Look at the elementary tree of arrives in the analysis of (8).)