An Incremental Parser for Abstract Meaning Representation 1 Marco - - PowerPoint PPT Presentation

an incremental parser for abstract meaning representation
SMART_READER_LITE
LIVE PREVIEW

An Incremental Parser for Abstract Meaning Representation 1 Marco - - PowerPoint PPT Presentation

An Incremental Parser for Abstract Meaning Representation 1 Marco Damonte 1 Shay B. Cohen 2 Giorgio Satta 1 University of Edinburgh 2 University of Padua EACL 2017 1 / 25 AMR He described her as a genius describe-01 ARG2 ARG1 ARG0 genius


slide-1
SLIDE 1

An Incremental Parser for Abstract Meaning Representation

1 Marco Damonte 1 Shay B. Cohen 2 Giorgio Satta 1 University of Edinburgh 2 University of Padua

EACL 2017

1 / 25

slide-2
SLIDE 2

AMR

He described her as a genius describe-01 he genius she ARG0 ARG1 ARG2

2 / 25

slide-3
SLIDE 3

Dependency trees

  • Transition-based dependency parsing (Nivre, 2004)

My dog also likes eating

root poss nsubj advmod xcomp

3 / 25

slide-4
SLIDE 4

Concept identification

The teacher person teach-01 ARG0-of The proposal thing propose-01 ARG1-of 10 January 1989 date-entity 10 1 1989 day month year

4 / 25

slide-5
SLIDE 5

Reentrancy

I beg you to excuse me beg-01 i you excuse-01 ARG0 ARG1 ARG2 ARG0 ARG1

5 / 25

slide-6
SLIDE 6

Reentrancy

I beg you to excuse me beg-01 i you excuse-01 ARG0 ARG1 ARG2 ARG0 ARG1

6 / 25

slide-7
SLIDE 7

Reentrancy

I beg you to excuse me beg-01 i you excuse-01 ARG0 ARG1 ARG2 ARG0 ARG1

7 / 25

slide-8
SLIDE 8

Transition-based AMR Parser

8 / 25

slide-9
SLIDE 9

Transition system

The boy wants to believe the girl STACK boy GRAPH

9 / 25

slide-10
SLIDE 10

Transition system

The boy wants to believe the girl STACK want-01 boy GRAPH want-01 boy

10 / 25

slide-11
SLIDE 11

Transition system

The boy wants to believe the girl STACK want-01 GRAPH want-01 boy

11 / 25

slide-12
SLIDE 12

Transition system

The boy wants to believe the girl STACK want-01 believe-01 GRAPH want-01 boy believe-01

12 / 25

slide-13
SLIDE 13

Transition system

The boy wants to believe the girl STACK want-01 believe-01 girl GRAPH want-01 boy believe-01 girl

13 / 25

slide-14
SLIDE 14

Transition system

The boy wants to believe the girl STACK want-01 believe-01 GRAPH want-01 boy believe-01 girl

14 / 25

slide-15
SLIDE 15

Transition system

The boy wants to believe the girl STACK want-01 believe-01 GRAPH want-01 boy believe-01 girl

15 / 25

slide-16
SLIDE 16

Transition system

The boy wants to believe the girl STACK want-01 GRAPH want-01 boy believe-01 girl

16 / 25

slide-17
SLIDE 17

Transition system

The boy wants to believe the girl STACK GRAPH want-01 boy believe-01 girl

17 / 25

slide-18
SLIDE 18

Oracle

  • Given the current configuration (σ, β, A) and the gold-standard

graph G = (Vg, Ag): T(G, σ, β, A) =                LARC(ℓ) RARC(ℓ) RED-REENT(ℓ) REDUCE SHIFT

  • (English, AMR) ⇒ Transitions to obtain AMR* from English

18 / 25

slide-19
SLIDE 19

Evaluation

19 / 25

slide-20
SLIDE 20

Fine-grained evaluation

  • Smatch. Cai and Knight (2013)
  • Unlabeled. Smatch score after removing edge labels
  • No WSD. Smatch score while ignoring Propbank senses
  • Reentrancy. Smatch computed on reentrant edges
  • Semantic Role Labelling. Smatch computed on :ARG roles

20 / 25

slide-21
SLIDE 21

Fine-grained evaluation

  • Smatch. Cai and Knight (2013)
  • Unlabeled. Smatch score after removing edge labels
  • No WSD. Smatch score while ignoring Propbank senses
  • Reentrancy. Smatch computed on reentrant edges
  • Semantic Role Labelling. Smatch computed on :ARG roles

20 / 25

slide-22
SLIDE 22

Fine-grained evaluation

  • Smatch. Cai and Knight (2013)
  • Unlabeled. Smatch score after removing edge labels
  • No WSD. Smatch score while ignoring Propbank senses
  • Reentrancy. Smatch computed on reentrant edges
  • Semantic Role Labelling. Smatch computed on :ARG roles

20 / 25

slide-23
SLIDE 23

Fine-grained evaluation

  • Smatch. Cai and Knight (2013)
  • Unlabeled. Smatch score after removing edge labels
  • No WSD. Smatch score while ignoring Propbank senses
  • Reentrancy. Smatch computed on reentrant edges
  • Semantic Role Labelling. Smatch computed on :ARG roles

20 / 25

slide-24
SLIDE 24

Fine-grained evaluation

  • Smatch. Cai and Knight (2013)
  • Unlabeled. Smatch score after removing edge labels
  • No WSD. Smatch score while ignoring Propbank senses
  • Reentrancy. Smatch computed on reentrant edges
  • Semantic Role Labelling. Smatch computed on :ARG roles

20 / 25

slide-25
SLIDE 25

Fine-grained evaluation (cont’d)

  • Concepts. F-score on the concept identification task
  • Negations. F-score on :polarity roles
  • Named Entities. F-score on :name roles
  • Wikification. F-score on :wiki roles

21 / 25

slide-26
SLIDE 26

Fine-grained evaluation (cont’d)

  • Concepts. F-score on the concept identification task
  • Negations. F-score on :polarity roles
  • Named Entities. F-score on :name roles
  • Wikification. F-score on :wiki roles

21 / 25

slide-27
SLIDE 27

Fine-grained evaluation (cont’d)

  • Concepts. F-score on the concept identification task
  • Negations. F-score on :polarity roles
  • Named Entities. F-score on :name roles
  • Wikification. F-score on :wiki roles

21 / 25

slide-28
SLIDE 28

Fine-grained evaluation (cont’d)

  • Concepts. F-score on the concept identification task
  • Negations. F-score on :polarity roles
  • Named Entities. F-score on :name roles
  • Wikification. F-score on :wiki roles

21 / 25

slide-29
SLIDE 29

Experiments

Metric JAMR (’14) CAMR JAMR (’16) Ours Smatch 58 63 67 64 Unlabeled 61 69 69 69 No WSD 58 64 68 65 NP-only 47 54 58 55 Reentrancy 38 41 42 41 Concepts 79 80 83 83 Named Ent. 75 75 79 83 Wikification 75 64 Negations 16 18 45 48 SRL 55 60 60 56 JAMR: Flanigan et al. (2014) CAMR: Wang et al. (2015)

22 / 25

slide-30
SLIDE 30

Software

  • Online demo:

http://cohort.inf.ed.ac.uk/amreager.html

  • Source code for parser:

https://github.com/mdtux89/amr-eager

  • Source code for evaluation:

https://github.com/mdtux89/amr-evaluation

23 / 25

slide-31
SLIDE 31

Demo

24 / 25

slide-32
SLIDE 32

Conclusions

  • AMREager is a linear-time, left-to-right transition system
  • AMR parsing akin to dependency parsing
  • Fine-grained evaluation suite to assess AMR parsers

25 / 25