Semantic Parsing based on Propositional Representations

1
Semantic Parsing based on Propositional
Representations

• Fernandez, Sopena, Padró

2
Generalized Role Labeling using Propositional
Representations

• Resumen
• Objetivos
• Estructura Proposicional
• Arquitectura
• M1
• M2
• Ejemplos
• Canónico
• PP-Attachment

3
Generalized Role Labeling using Propositional
Representations

• Objetivos
• Crear un analizador semántico mediante la
  implementación de un modelo psicológico plausible
  que
• Lleva a cabo un mapeo directo i sencillo de las
  frases a su estructura proposicional
• No utiliza analizadores sintácticos ni estructura
  sintactica intermedia
• Obtener buenos resultados en textos reales (PTB)

4
Generalized Role Labeling using Propositional
Representations

• Estructura proposicional
• Predicado 3 Argumentos
• Frase canónica The man sold some offerings to
  the british tourist
• Predsold
• Arg1the man
• Arg2some offerings
• Arg3the british tourist
• Composición de proposiciones
• The man sold some offerings to the british
  tourist in Barcelona

(P1) Predsold Arg1the man Arg2some
offerings Arg3the british tourist
(P2) Pred Arg1P1 Arg2in Barcelona Arg3
5
Generalized Role Labeling using Propositional
Representations
6
Generalized Role Labeling using Propositional
Representations

• Estructura proposicional
• Argumentos temáticos generalizados (VanValin)
• A1-ACTOR (agent, perceiver, ....)
• A2-UNDERGOER (theme, patient, ... )
• A3-OTHERS (benefactive, goal, location, source,
  destination, ...)
• Mapping fácil con FrameNet i otros
• Arg1 i arg2 son los dos primeros argumentos core
• Arg3, argumentos core que se identifican por la
  preposición que los marca.

7
Generalized Role Labeling using Propositional
Representations

• Arquitectura
• Modulo1 - Estructural/Sintáctico
• Lleva a cabo el mapeo directo de las palabras a
  la proposición
• Modifica la proposición
• Sin información semántica explicita
• Modulo2 - Semántico
• Acepta o rechaza las decisiones estructurales del
  primer modulo.
• Consistencia con el verbo (/-Subcategorization
  Frames)
• PP-Attachment
• Coordinacion
• Relativo

8
Ventajas

• Complejidad lineal
• Fácil tratamiento de fenómenos sintácticos
  difíciles
• Coordinación y puntuación.
• Word order
• Non local dependencies.
• No se necesita corpus sintáctico etiquetado.

9
Parser Commands

• Arquitectura
• Modulo1

STACK
Stored Context
Input Word
Slot 0 Slot 1 Slot 2 Slot 3 Type S
Back, Test Subcat.

Current Context
10
VerboARG
AdjuntoAdjunto
CoordinableCoordinable
Arg1Arg2
Input Word
Slot 0 Slot 1 Slot 2 Slot 3 Type S
Back, Test Subcat.

Current Context
11
PUT1
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
Slot 3
Flags
12
The man sold some offerings to the president
The DT
PUT1
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
Slot 3
Flags
13
The man sold some offerings to the president
The DT
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
DT(The)
Slot 3
Flags
14
The man sold some offerings to the president
man DT_N
PUT1
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
DT(The)
Slot 3
Flags
15
The man sold some offerings to the president
man DT_N
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
DT(The) DT_N(man)
Slot 3
Flags
16
The man sold some offerings to the president
sold V_MA
PUT0
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
DT(The) DT_N(man)
Slot 3
Flags
17
The man sold some offerings to the president
sold V_MA
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
V_MA(sold)
DT(The) DT_N(man)
Slot 3
Flags
18
The man sold some offerings to the president
some DT
PUT2
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
V_MA(sold)
DT(The) DT_N(man)
Slot 3
Flags
19
The man sold some offerings to the president
some DT
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
V_MA(sold)
DT(The) DT_N(man)
DT(some)
Slot 3
Flags
20
The man sold some offerings to the president
offerings DT_N
PUT2
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
V_MA(sold)
DT(The) DT_N(man)
DT(some)
Slot 3
Slot 3
Flags
21
The man sold some offerings to the president
offerings DT_N
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
V_MA(sold)
DT(The) DT_N(man)
DT(some) DT_N(offerings)
Slot 3
Slot 3
Flags
22
The man sold some offerings to the president
to IIN_DT
PUT3
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
V_MA(sold)
DT(The) DT_N(man)
DT(some) DT_N(offerings)
Slot 3
Slot 3
Flags
23
The man sold some offerings to the president
to IIN_DT
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
V_MA(sold)
DT(The) DT_N(man)
DT(some) DT_N(offerings)
to(IIN_DT)
Slot 3
Flags
24
The man sold some offerings to the president
the DT
PUT1
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
V_MA(sold)
DT(The) DT_N(man)
DT(some) DT_N(offerings)
to(IIN_DT)
Slot 3
Flags
25
The man sold some offerings to the president
the DT
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
V_MA(sold)
DT(The) DT_N(man)
DT(some) DT_N(offerings)
to(IIN_DT) the(DT)
Slot 3
Slot 3
Flags
26
The man sold some offerings to the president
president DT_N
PUT1
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
V_MA(sold)
DT(The) DT_N(man)
DT(some) DT_N(offerings)
to(IIN_DT) the(DT)
Slot 3
Slot 3
Flags
27
The man sold some offerings to the president
president DT_N
Modul1
Modul2
Slot 0
Slot 1
Slot 2
Slot 3
V_MA(sold)
DT(The) DT_N(man)
DT(some) DT_N(offerings)
to(IIN_DT) the(DT)
Slot 3
Slot 3
president(DT_N)
Flags
28
Generalized Role Labeling using Propositional
Representations
Un ejemplo no tan sencilloThe main manager
bought some old cars with three wheels.
29
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word The DT M1 PUT1 M2
Current Pred A1 A2 A3 Flags
30
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word The DT M1 NEXT M2
Current Pred A1 The A2 A3 Flags _at_1
31
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word main JJ_PR M1 IZ-IN M2
Current Pred A1 The A2 A3 Flags _at_NEXT _at_1
32
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word main JJ_PR M1 PUT0 M2
Current Pred A1 A2 A3 Flags
Top Pred A1 The A2 A3 Flags _at_1
33
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word main JJ_PR M1 NEXT M2
Current Pred main A1 A2 A3 Flags
Top Pred A1 The A2 A3 Flags _at_1
34
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word manager DT_N M1 PUT1 M2
Current Pred main A1 A2 A3 Flags _at_NEXT
Top Pred A1 The A2 A3 Flags _at_1
35
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word manager DT_N M1 OZ-OUT M2
Current Pred main A1 manager A2 A3 Flags
_at_NEXT
Top Pred A1 The A2 A3 Flags _at_1
36
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word manager DT_N M1 PUT1 M2
PmainA1manager
Current Pred A1 The A2 A3 Flags _at_1 _at_OZ-OUT
37
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word manager DT_N M1 NEXT M2
PmainA1manager
Current Pred A1 The manager A2 A3 Flags _at_1
_at_OZ-OUT
38
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word bought V_MA M1 PUT0 M2
PmainA1manager
Current Pred A1 The manager A2 A3 Flags _at_1
_at_NEXT
39
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word bought V_MA M1 NEXT M2
PmainA1manager
Current Pred bought A1 The manager A2 A3 Flag
s _at_0
40
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word some DT M1 PUT2 M2
PmainA1manager
Current Pred bought A1 The manager A2 A3 Flag
s _at_0 _at_NEXT
41
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word some DT M1 NEXT M2
PmainA1manager
Current Pred bought A1 The manager A2
some A3 Flags _at_2
42
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word old JJ_PR M1 IZ-IN M2
PmainA1manager
Current Pred bought A1 The manager A2
some A3 Flags _at_2 _at_NEXT
43
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word old JJ_PR M1 PUT0 M2
PmainA1manager
Current Pred A1 A2 A3 Flags _at_IZ-IN
Top Pred bought A1 The manager A2
some A3 Flags _at_2
44
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word old JJ_PR M1 NEXT M2
PmainA1manager
Current Pred old A1 A2 A3 Flags _at_IZ-IN _at_0
Top Pred bought A1 The manager A2
some A3 Flags _at_2
45
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word cars DT_N M1 PUT1 M2
PmainA1manager
Current Pred old A1 A2 A3 Flags _at_IZ-IN
_at_NEXT _at_0
Top Pred bought A1 The manager A2
some A3 Flags _at_2
46
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word cars DT_N M1 OZ-OUT M2
PmainA1manager
Current Pred old A1 cars A2 A3 Flags _at_IZ-IN
_at_NEXT _at_1
Top Pred bought A1 The manager A2
some A3 Flags _at_2
47
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word cars DT_N M1 PUT2 M2
PmainA1manager
PoldA1cars
Current Pred bought A1 The manager A2
some A3 Flags _at_2
48
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word cars DT_N M1 NEXT M2
PmainA1manager
PoldA1cars
Current Pred bought A1 The manager A2 some
cars A3 Flags _at_2
49
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word with IIN_DT M1 IZ-IN1 M2
PmainA1manager
PoldA1cars
Current Pred bought A1 The manager A2 some
cars A3 Flags _at_2 _at_NEXT
50
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word with IIN_DT M1 PUT2 M2
PmainA1manager
PoldA1cars
Current Pred A1 cars A2 A3 Flags _at_1
Top Pred bought A1 The manager A2 some
cars A3 Flags _at_2
51
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word with IIN_DT M1 NEXT M2
PmainA1manager
PoldA1cars
Current Pred A1 some cars A2 with A3 Flags
_at_2
Top Pred bought A1 The manager A2 some
cars A3 Flags _at_2
52
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word three DT M1 PUT2 M2
PmainA1manager
PoldA1cars
Current Pred A1 some cars A2 with A3 Flags
_at_2 _at_NEXT
Top Pred bought A1 The manager A2 some
cars A3 Flags _at_2
53
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word three DT M1 NEXT M2
PmainA1manager
PoldA1cars
Current Pred A1 some cars A2 with
three A3 Flags _at_2
Top Pred bought A1 The manager A2 some
cars A3 Flags _at_2
54
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word wheels DT_N M1 PUT2 M2
PmainA1manager
PoldA1cars
Current Pred A1 some cars A2 with
three A3 Flags _at_2 _at_NEXT
Top Pred bought A1 The manager A2 some
cars A3 Flags _at_2
55
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word wheels DT_N M1 NEXT M2
PmainA1manager
PoldA1cars
Current Pred A1 some cars A2with three
wheels A3 Flags _at_2
Top Pred bought A1 The manager A2 some
cars A3 Flags _at_2
56
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word . DOT M1 OZ-OUT M2
PmainA1manager
PoldA1cars
Current Pred A1 some cars A2with three
wheels A3 Flags _at_2 _at_NEXT
Top Pred bought A1 The manager A2 some
cars A3 Flags _at_2
57
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word . DOT M1 OZ-OUT M2
PmainA1manager
PoldA1cars
A1carsA2with three wheels
Current Pred bought A1 The manager A2 some
cars A3 Flags _at_2
58
Generalized Role Labeling using Propositional
Representations

• The main manager bought some old cars with three
  wheels.

Input Word . DOT M1 FIN M2
PmainA1manager
PoldA1cars
A1carsA2with three wheels
PboughtA1the manager A2some cars
59
Training set

• SS-3-2- (DT_N I PUT1 NEXT)
• SS-3-2- (V_MA took PUT0 NEXT)
• SS-3-3- (DT_N her PUT2 NEXT)
• SS-3-4- (IIN_DT to PUT3 NEXT)
• SS-3-5- (DT_N school PUT3 NEXT)
• SS-3-6- (. . OZ-OUT NEXT)
• SS-3-7- (FIN)

60
Elementary expressions

• The non-callable issue, which can be put_back to
  the company in 1999, is priced at 99 basis_points
  above the Treasury 10-year note.

Current Pred non-callable A1 issue A2 A3 Flags

61
Elementary expressions

• The non-callable issue, which can be put_back to
  the company, is priced at 99 basis_points above
  the Treasury 10-year note.

Current Predcan be put_back A1 A2 the
issue A3 to the company
62
Elementary expressions

• The non-callable issue, which can be put_back to
  the company, is priced at 99 basis_points above
  the Treasury 10-year note.

Current Predis priced A1 A2 the issue A3 at
99 basis_points
63
Elementary expressions

• The non-callable issue, which can be put_back to
  the company, is priced at 99 basis_points above
  the Treasury 10-year note.

Current Pred A1 99 basis_points A2 above
the A3
64
Elementary expressions

• The non-callable issue, which can be put_back to
  the company, is priced at 99 basis_points above
  the Treasury 10-year note.

Current Pred A1 A2 treasury A3
65
Elementary expressions

• The non-callable issue, which can be put_back to
  the company, is priced at 99 basis_points above
  the Treasury 10-year note.

Current Pred A1 note A2 10-year A3
66
Elementary expressions

• The non-callable issue, which can be put_back to
  the company, is priced at 99 basis_points above
  the Treasury 10-year note.

Current Pred A1 note A2 Treasury A3
67
Elementary expressions

• The non-callable issue, which can be put_back to
  the company, is priced at 99 basis_points above
  the Treasury 10-year note.

Current Predis priced A1 A2 the issue A3 at
99 basis_points
68
Training set

• SS-17-1- (DT_N IMB PUT1 NEXT)
• SS-17-2- (V_MA bought PUT0 NEXT)
• SS-17-3- (DT the PUT2 NEXT)
• SS-17-4- (DT_N team PUT2 NEXT)
• SS-17-5- (IIN_DT from PUT3 NEXT)
• SS-17-6- (DT_N BUMBRIGHT PUT3 NEXT)
• SS-17-7- (IIN_DT for CONSTRUAL PUT2 NEXT)
• SS-17-8- (DT 145.000 PUT2 BACK OZ-OUT PUT3 NEXT)
• SS-17-9- (DT_N PUT3 NEXT)
• SS-17-10- (. . OZ-OUT NEXT)
• SS-17-11- (FIN)

69
FIN