Transcript Ling 001: Syntax II

Ling 001: Syntax II
Rules, Movement, Ambiguity
Phrases
• In the last lecture, we talked about simple
phrases; e.g. Noun Phrases like
– The dog
– The big dog
– The big dog that John was talking to
• In this lecture, we will look at
– how phrases and larger objects are derived by
rules, and
– how phrases can be moved from one position to
another
• We will also look at structural ambiguity
Reviewing….
• Remember that in the last lecture we developed
some basic notions about constituency.
• Let’s apply these to sentences. Consider:
The boy kicked the ball
• We have three lexical categories here; the nouns boy,
ball, and the verb kick
• This gives us three phrases: two NPs, (subject and
object), and one VP (headed by kick)
• Determining how these phrases are organized into
the sentence involves the same reasoning we applied
above
Possible structures
• In principle, the three phrases could be arranged in
two ways; this is exactly parallel to what we did with
words before (I’m using ‘S’ here as the label for
‘sentence’):
Structure 1
Structure 2
S
S
VP
NP
the boy
V
NP
kicked the ball
VP
NP
the boy
V
kicked
NP
the ball
The options
• The different structures take different
positions on the status of the VP; is it
– The object and verb that form a VP, or
– The subject and verb that form a VP?
• We can use the diagnostics above to give us
an answer
Tests
Substitution:
– John ate an apple.
– Mary did too.
– Did = <ate an apple>
 Verb + Object behaves like a constituent
Movement:
Mary said she would fix the car with a wrench
…and [fix the car with a wrench] she did
 Tests indicate that Verb + Object behave like a constituent
(structure 1)
Rules…
• The review we just did gives us a natural
transition to our next topic
• Recall that one of the things that we have to
account for in syntactic theory is how
language makes infinite use of a finite
number of words
• We’ll see how this can be done using a basic
grammar. Although our grammar will be a toy,
even simple tools like this suffice to illustrate
the main point
What we need
• We’re going to assume, as we’ve said, that there are
– A set of words, which belong to different categories; and
– A set of rules, that account for how phrases are built
• For the first part, let’s take some nouns, verbs,
determiners, and adjectives:
– Nouns: cat, boy, book, burrito
– Verbs(transitive): eats, reads, pets
– Determiners: a, the
– Adjectives: orange, small, stubborn, purring
• These words will be the bottom elements (terminals)
in our syntactic trees
Some Rules
• We can use simple phrase structure rules to give us
the basic trees we will work with
• Remember above that the verb forms a constituent
with the object; this suggests that our starting rule
should be:
S --> NP VP
• Semi-formally, the symbol S (for “sentence”) is
expanded into a Noun Phrase and a Verb Phrase
• We need to keep working until we have all terminals
(words); this means we need rules for expanding the
VP and the NP
Verb Phrase
• For our example, take the following rule for the VP:
VP --> V NP
I.e., a Verb Phrase expands into a Verb and a Noun
Phrase
• We now have two noun phrases to deal with
• For the V part, take
V --> {eats, reads, pets}
• When we reach one of these words for the V, we’re
done with that branch
Noun Phrases
• Now, let’s talk about Noun Phrases (NPs). These have (among
other properties) the following:
– The optional presence of a determiner (‘the’, ‘a’, etc.)
– The optional presence of more than one adjective
• We can write a rule that generates NPs in the following way:
NP --> (det) AP* N
This means that a noun phrase consists of minimally a head N; it
also can have
-an optional determiner (parentheses)
-any number of Adjective Phrases (AP), including zero; assume
AP --> A
• From this rule, and rules that say ‘N-->cat,…’, A --> ‘orange,
small, stubborn, purring …’, we can generate a number of
phrases
Examples
• From N --> (det) AP* N
NP
det
the
AP
AP
A
A
N
small purring cat
Lots of sentences
• Even with the limited vocabulary we have, and these
simple rules, we generate a lot of sentences; e.g.
– A small boy pets the orange cat.
– The orange cat eats the small burrito.
– The stubborn boy pets the orange book.
• And many others (think about it…)
• Two additional points
– One about sentences that are ungrammatical/make no
sense
– One about making the grammar put sentences within
sentences
First…
• Two types of sentences generated by our grammar
that are deviant
– Type 1 (morpho-syntactic?):
• *Boy reads the orange book.
• *The small cat eats book.
• *The cat eats a orange burrito.
– Type 2 (semantic?)
• The purring orange burrito reads a stubborn
boy.
– Many, many more like Type 2. Also some unclear cases; e.g.
The orange stubborn cat…
• As you can see, such a grammar generates a
number of sentences that are deviant in some way.
We either need to fix the grammar so that it doesn’t
do this, or attribute the deviance to e.g. semantics.
Second…
• Our rules above do not allow us to have sentences
within sentences.
• A simple change fixes this:
VP --> V+ CP
Where
– V+ --> {say, know, that}; and
– CP --> Comp S
– Comp --> {that}
• With this addition, VPs can have verbs that take a
sentence inside of them
• It is now possible to derive The cat knows that a boy
said that the burrito thinks that….
Interim
• As you can see, simple grammars can be extremely
powerful. Syntactic theory tries to get the right
balancer of power and constraints to account for
people’s syntactic competence.
• In our examples, above, we derived basic syntactic
structures
• A complication to this picture is found in the fact that
languages seem to move constituents from one place
to another
• (Remember our first discussion of question
formation)
Arguments, etc.
• In some sense, many things that happen in a
sentence depend on what the verb in the
sentence is:
– Transitive verb: kick
• Two ‘arguments’ of kick
– Intransitive verb: sleep
• One argument
• In order to be more precise about this, we
need to distinguish grammatical (syntactic)
position from semantic role
Roles and Positions
• Consider a transitive verb like kick
– This has two arguments
– The arguments are
• The agent (the kicker)
• The patient (the thing kicked)
– In normal, active sentences in English
• The agent is the subject
• The patient is the object
– There are not verbs blick like kick where this
relationship is reversed
Further points
• In the examples above, there is a correspondence:
Agent/Subject, Patient/Object
• One case where this is not found is in passive
sentences
• Consider
– The boy kicked the ball
– The ball was kicked by the boy
• In the passive, the Agent and the Patient are the
same as they are in the active
• But the syntactic positions are different; in particular
– The Patient of the verb is in the Subject position
– The subject in English appears in a specific position in the
clause, and e.g. controls agreement on the verb
Verbs and Arguments
• Verbs are looking for their arguments in particular
positions; remember the rule we formulated above:
– Patients appear in object position (inside the VP)
• What about the passive? Here is where the process
of movement is important
• We can start with the VP [kick [the ball]].
• Then the object of the verb kick is moved to subject position as
part of the passive rule
• It is still interpreted as the Patient, because that is where it starts
• In order to be interpreted as a patient, the NP has to have some
relationship to the position where it came from; this is why
traces are important:
• [The ball] was kicked t(race)
Other cases of movement
• The same principle applies in other areas as well:
• Questions
– John ate the apples.
– What did John eat t
• Relative clauses
– John was talking to Mary.
– The woman [who John was talking to t]
• Topicalization
– John likes these apples.
– These apples John likes t.
Movement, cont.
• What does movement allow us to say?
– Uniformly, verbs look for their arguments in particular places
– These arguments are interpreted by fixed rules; e.g. ‘objects
are patients’
– Even when the surface order of constituents does not match
this underlying design, the same rules apply (as long as we
have traces)
– Sometimes the original structure is called D-structure, while
the derived structure is called S-structure (think ‘deep’ vs.
‘surface’:
– Example:
• D-Structure: John ate what
• S-Structure: What did John eat t?
Other examples
• In the examples above, it’s mostly objects that are
moved. But this is not always the case; consider:
– John said Bill fixed the car.
– Who did John say t fixed the car?
• Or
– John fixed the car [with a wrench].
– How did John fix the car t?
• We’ll try to use simple examples when asking you to
find traces. Remember, the best method is to put
things “back into the regular order”, this should allow
you to see where the trace is…
(Structural) Ambiguities
• Notice that both NPs and VPs can have PPs
attached to them
• In some cases, this results in what is called a
structural ambiguity: one string has more
than one structure associated with it, and
means different things depending on what the
structure is
• Example: John saw the man with the
telescope
– Reading 1: John used a telescope
– Reading 2: The man John saw had a telescope
(Structural) Ambiguities
• Notice that both NPs and VPs can have PPs
attached to them
• We could do this with rules like VP --> VP PP, NP -->
NP PP
• In some cases, this results in what is called a
structural ambiguity: one string has more than one
structure associated with it, and means different
things depending on what the structure is
• Example: John saw the man with the telescope
– Reading 1: John used a telescope
– Reading 2: The man John saw had a telescope