Syntactic Theory When displaced did it leave a trace? Clayton - - PowerPoint PPT Presentation

syntactic theory
SMART_READER_LITE
LIVE PREVIEW

Syntactic Theory When displaced did it leave a trace? Clayton - - PowerPoint PPT Presentation

Syntactic Theory When displaced did it leave a trace? Clayton Greenberg Department of Language Science and Technology, Saarland University 8 December 2016 C. Greenberg (UdS LST) Traces 8 December 2016 1 / 26 Introduction How to read these


slide-1
SLIDE 1

Syntactic Theory

When displaced did it leave a trace? Clayton Greenberg

Department of Language Science and Technology, Saarland University

8 December 2016

  • C. Greenberg (UdS LST)

Traces 8 December 2016 1 / 26

slide-2
SLIDE 2

Introduction

How to read these slides

Green: important terms Blue: definitions Blue ≈: approximate definitions Purple: examples Italic naturally unpronounced elements Strike-out elements that were deleted

  • C. Greenberg (UdS LST)

Traces 8 December 2016 2 / 26

slide-3
SLIDE 3

Introduction

Displacement is...

  • “displacement is the capability of language to communicate about things

that are not immediately present (spatially or temporally); i.e., things that are either not here or are not here now.” (Wikipedia) “displacement is the capability of language to communicate about things that are not immediately present (spatially or temporally); i.e., things that are either not here or are not here now.” (Wikipedia) Very nice, but this is not the displacement I was looking for.

  • “alternations in word order”. Examples: particle verb alternation, adverb

preposing, clefting, extraposition“alternations in word order”. Examples: verb-particle alternation, adverb preposing, clefting, extraposition

  • This is almost the displacement I was looking for, but not quite.
  • Displacement is the phenomenon in which a constituent is uttered in a

position other than where it is logically interpreted.

  • C. Greenberg (UdS LST)

Traces 8 December 2016 3 / 26

slide-4
SLIDE 4

Reasons to displace

Reasons to displace

  • To get Case
  • To “check” a feature, such as +Q
  • To satisfy a dependent morpheme
  • (Optionally) to focus a particular participant or event
  • C. Greenberg (UdS LST)

Traces 8 December 2016 4 / 26

slide-5
SLIDE 5

Reasons to displace

The template tree

TP NP T′ T VP NP V′ V CP NP C′ C TP NP T′ T VP NP V′ V CP

  • C. Greenberg (UdS LST)

Traces 8 December 2016 5 / 26

slide-6
SLIDE 6

Reasons to displace Displace for Case

Passivization

TP NP D the N crook T′ T was VP NP PRO V′ V arrested NP D the N crook

  • C. Greenberg (UdS LST)

Traces 8 December 2016 6 / 26

slide-7
SLIDE 7

Reasons to displace Displace for Case

Seems

“seems” has a special argument structure:

  • nly one θ-role in

V-Comp position So, no need for a spec-VP position

TP NP D the N dog T′ T

  • pres

VP V seems CP C

/

TP NP D the N dog T′ T to VP NP D the N dog V′ V want NP bacon

  • C. Greenberg (UdS LST)

Traces 8 December 2016 7 / 26

slide-8
SLIDE 8

Reasons to displace Displace for feature

Head Movement from T to C (English)

TP NP I T′ T don’t VP NP I V′ V know CP C whether TP NP he T′ T will VP V be VP NP he V′ V reading NP D the N book

CP C will TP NP he T′ T will VP V be VP NP he V′ V reading NP D the N book

  • C. Greenberg (UdS LST)

Traces 8 December 2016 8 / 26

slide-9
SLIDE 9

Reasons to displace Displace for feature

Head Movement from T to C (Russian)

TP NP Ja T′ T ne VP NP Ja V′ V znagu CP C budet-li TP NP

  • n

T′ T budet VP V

  • prog

VP NP

  • n

V′ V ˇ citat’ NP knigu

CP C budet-li TP NP

  • n

T′ T budet VP V

  • prog

VP NP

  • n

V′ V ˇ citat’ NP knigu

  • C. Greenberg (UdS LST)

Traces 8 December 2016 9 / 26

slide-10
SLIDE 10

Reasons to displace Displace for feature

wh-movement with pied-piping

CP NP D Which N′ N student PP P

  • f

NP syntax C′ C

/

TP NP D Which N′ N student PP P

  • f

NP syntax T′ T is A late

  • C. Greenberg (UdS LST)

Traces 8 December 2016 10 / 26

slide-11
SLIDE 11

Reasons to displace Displace for feature

Logical forms

  • Spell-Out: where phonetic form (PF) and logical form (LF) diverge
  • PFs have no trees and LFs have no pronunciation (economy of

representation).

  • “Duplicates” must be deleted at PF (after Spell-Out) so they are not

pronounced multiple times.

  • A question at LF: quantifier, domain, predicate.

For which x, x a student of syntax, x was late

  • The original wh-phrase is a variable, bound by the copy at spec-CP

.

  • Chinese has wh-phrases in situ: the higher copy is deleted at PF
  • C. Greenberg (UdS LST)

Traces 8 December 2016 11 / 26

slide-12
SLIDE 12

Reasons to displace Displace for focus

Displace for focus: verb-particle alternation

  • ≈ Perhaps a string operation that occurs at PF (after Spell-Out)
  • I threw up my lunch.
  • I threw my lunch up.
  • C. Greenberg (UdS LST)

Traces 8 December 2016 12 / 26

slide-13
SLIDE 13

Reasons to displace Displace for focus

Displace for focus: adverb preposing

FocusP Focus P with NP D the N telescope TP NP D the N man T′ T

  • past

VP NP D the N man V′ V′ V saw NP D the N woman PP P with NP D the N telescope

  • C. Greenberg (UdS LST)

Traces 8 December 2016 13 / 26

slide-14
SLIDE 14

Reasons to displace Displace for focus

Displace for focus: clefting

  • It-cleft: It is Jaime for whom we are looking.
  • Wh-cleft/Pseudo-cleft: What he wanted to buy was a Fiat.
  • Reversed wh-cleft/Inverted pseudo-cleft: A Fiat is what he wanted to

buy.

  • All-cleft: All he wanted to buy was a Fiat.
  • Inferential cleft: It is not that he loves her. It’s just that he has a way

with her that is different.

  • There-cleft: And then there’s a new house he wanted to build.
  • If-because cleft: If he wants to be an actor it’s because he wants to be

famous.

  • C. Greenberg (UdS LST)

Traces 8 December 2016 14 / 26

slide-15
SLIDE 15

Reasons to displace Displace for focus

Displace for focus: it-clefting

  • NP: It was a cookie that she ate.
  • PP: It was on foot that he went there.
  • AdvP: It was greedily and speedily that Homer Simpson drank his beer.
  • Inf clause: It is to address a far-reaching problem that Oxfam is

launching this campaign.

  • Gerund: It could be going home early or slacking off work that the boss

reacted to.

  • Adv clause: It was because she was so lonely all the time that she

decided to move out.

  • C. Greenberg (UdS LST)

Traces 8 December 2016 15 / 26

slide-16
SLIDE 16

Reasons to displace Displace for focus

Displace for focus: extraposition

Someone who we don’t know left a message. Someone left a message who we don’t know. Susan said something that nobody expected more than once. Susan said something more than once that nobody expected. Some guy with red hair was there. Some guy was there with red hair. How frustrated with their kids are they? How frustrated are they with their kids? *What that was so entertaining actually happened? What actually happened that was so entertaining? *What that upset everyone do you think they did? What do you think they did that upset everyone?

  • C. Greenberg (UdS LST)

Traces 8 December 2016 16 / 26

slide-17
SLIDE 17

Evidence for traces

Wanna contraction

CP NP who C′ C does TP NP Vicky T′ T does VP NP Vicky V′ V want CP NP who C′ C

/

TP NP who T′ T to VP NP who V win

CP NP who C′ C does TP NP Vicky T′ T does VP NP Vicky V′ V want CP NP who C′ C / TP NP PRO T′ T to VP NP PRO V′ V vote PP P for NP who

  • C. Greenberg (UdS LST)

Traces 8 December 2016 17 / 26

slide-18
SLIDE 18

Evidence for traces

What blocks contraction?

  • Intervening overt elements
  • duplicate constituents that were reMerged, i.e. traces!
  • Not the silent complementizer
  • Not PRO
  • C. Greenberg (UdS LST)

Traces 8 December 2016 18 / 26

slide-19
SLIDE 19

Evidence for traces

Subjacency

TP NP it T′ T

  • pres

VP V seems CP C / TP NP it T′ T to VP V have VP V been V′ V reported CP C that TP NP Charles T′ T is AP A likely CP C / TP NP Charles T′ T to VP NP Charles V′ V help NP us

  • C. Greenberg (UdS LST)

Traces 8 December 2016 19 / 26

slide-20
SLIDE 20

Evidence for traces

What are we allowed to reMerge?

  • subjacency: cannot reMerge an element over more than one bounding

node at a time

  • Bounding nodes vary by language!
  • In English, the bounding nodes are NP and TP

.

  • In many other languages, CP is a bounding node.
  • C. Greenberg (UdS LST)

Traces 8 December 2016 20 / 26

slide-21
SLIDE 21

A case study in structure sharing

The whole sentence

seems It nsubj reported xcomp xsubj to aux have aux been auxpass likely ccomp ROOT root that mark Charles nsubj help xcomp is cop xsubj to aux us dobj

  • C. Greenberg (UdS LST)

Traces 8 December 2016 21 / 26

slide-22
SLIDE 22

A case study in structure sharing

The full alternative

seems Charles nsubj reported xcomp xsubj to aux have aux been auxpass likely xcomp ROOT root to aux be cop help xcomp to aux us dobj

  • C. Greenberg (UdS LST)

Traces 8 December 2016 22 / 26

slide-23
SLIDE 23

A case study in structure sharing

Drop a verb, try again

seems Charles nsubj likely xcomp xsubj to aux be cop help xcomp ROOT root to aux us dobj

  • C. Greenberg (UdS LST)

Traces 8 December 2016 23 / 26

slide-24
SLIDE 24

A case study in structure sharing

Drop a copula, try again

likely Charles nsubj help xcomp seems cop xsubj to aux us dobj ROOT root

  • C. Greenberg (UdS LST)

Traces 8 December 2016 24 / 26

slide-25
SLIDE 25

Conclusion

Classical parser complexity

  • Merge-style parsers are considered intractable
  • Basic PCFG chart parser: O(n3)
  • Basic non-projective dependency parser: O(n2)
  • Basic projective dependency parser: O(n)
  • How much do we need to scale? How complex is acceptable?
  • C. Greenberg (UdS LST)

Traces 8 December 2016 25 / 26

slide-26
SLIDE 26

Conclusion

Summary

1 Memorize:

a. Displacement : uttered position = interpret position b. Spell-Out : where LF and PF diverge c. A question at LF : quantifier, domain, predicate d. Subjacency : don’t cross > 1 bounding domain

2 Traces can be handled using copies or structure sharing. 3 Copies respect linear order, but structure sharing seems less

computationally complex.

4 To play with the Stanford constituency and dependency parsers, see

http://eztreesee.coli.uni-saarland.de

5 To play with a minimalist parser, see

http://elmo.sbs.arizona.edu/sandiway/mpp/mm.html

  • C. Greenberg (UdS LST)

Traces 8 December 2016 26 / 26