An Extended GHKM Algorithm for Inducing λ -SCFG

32
An Extended GHKM Algorithm for Inducing λ-SCFG Peng Li [email protected] Tsinghua University

description

An Extended GHKM Algorithm for Inducing λ -SCFG. Peng Li [email protected] Tsinghua University. Semantic Parsing. Mapping natural language (NL) sentence to its computable meaning representation (MR). NL: Every boy likes a star. MR:. predicate. variable. Motivation. - PowerPoint PPT Presentation

Transcript of An Extended GHKM Algorithm for Inducing λ -SCFG

Page 1: An Extended GHKM Algorithm for Inducing  λ -SCFG

An Extended GHKM Algorithm forInducing λ-SCFG

Peng [email protected] University

Page 2: An Extended GHKM Algorithm for Inducing  λ -SCFG

Semantic Parsing

• Mapping natural language (NL) sentence to its computable meaning representation (MR)

NL: Every boy likes a star

MR:

variablepredicate

Page 3: An Extended GHKM Algorithm for Inducing  λ -SCFG

Motivation

• Common way: inducing probabilistic grammarPCFG: Probabilistic Context Free Grammar

Page 4: An Extended GHKM Algorithm for Inducing  λ -SCFG

Motivation

• Common way: inducing probabilistic grammarCCG: Combinatory Categorial Grammar

Page 5: An Extended GHKM Algorithm for Inducing  λ -SCFG

Motivation

• Common way: inducing probabilistic grammarSCFG: Synchronous Context Free Grammar

Page 6: An Extended GHKM Algorithm for Inducing  λ -SCFG

Motivation

• State of the art: SCFG + λ-calculus (λ-SCFG)• Major challenge: grammar induction– It is much harder to find the correspondence between

NL sentence and MR than between NL sentences• SCFG rule extraction is well-studied in MT• GHKM is the most widely used algorithm• We want to adapt GHKM to semantic parsing• Experimental results show that we get the state-

of-the-art performance

Page 7: An Extended GHKM Algorithm for Inducing  λ -SCFG

Background

• State of the art: SCFG + λ-calculus (λ-SCFG)

• λ-calculus– λ-expression: – β-conversion: bound variable substitution

– α-conversion: bound variable renaming

Page 8: An Extended GHKM Algorithm for Inducing  λ -SCFG

λ-SCFG Rule Extraction

• Outline1. Building training examples

1. Transforming logical forms to trees2. Aligning trees with sentences

2. Identifying frontier nodes3. Extracting minimal rules4. Extracting composed rules

Page 9: An Extended GHKM Algorithm for Inducing  λ -SCFG

Building Training Examples

NL: Every boy likes a star

MR:

Page 10: An Extended GHKM Algorithm for Inducing  λ -SCFG

Building Training Examples

Page 11: An Extended GHKM Algorithm for Inducing  λ -SCFG

Building Training Examples

Page 12: An Extended GHKM Algorithm for Inducing  λ -SCFG

Building Training Examples

boy human pop like

Page 13: An Extended GHKM Algorithm for Inducing  λ -SCFG

Building Training Examplesboy human pop like

Every boy likes a star

Page 14: An Extended GHKM Algorithm for Inducing  λ -SCFG

Identifying Frontier Nodes

Page 15: An Extended GHKM Algorithm for Inducing  λ -SCFG

Identifying Frontier Nodes

Page 16: An Extended GHKM Algorithm for Inducing  λ -SCFG

Identifying Frontier Nodes

Page 17: An Extended GHKM Algorithm for Inducing  λ -SCFG

Identifying Minimal Frontier Tree

Page 18: An Extended GHKM Algorithm for Inducing  λ -SCFG

Identifying Minimal Frontier Tree

Page 19: An Extended GHKM Algorithm for Inducing  λ -SCFG

Identifying Minimal Frontier Tree

Page 20: An Extended GHKM Algorithm for Inducing  λ -SCFG

Identifying Minimal Frontier Tree

Page 21: An Extended GHKM Algorithm for Inducing  λ -SCFG

Minimal Rule Extraction

X

X

Page 22: An Extended GHKM Algorithm for Inducing  λ -SCFG

Minimal Rule Extraction

X

X

Page 23: An Extended GHKM Algorithm for Inducing  λ -SCFG

Minimal Rule Extraction

X

X

Page 24: An Extended GHKM Algorithm for Inducing  λ -SCFG

Composed Rule Extraction

Page 25: An Extended GHKM Algorithm for Inducing  λ -SCFG

λ-SCFG Rule Extraction

• Outline1. Building training examples

1. Transforming logical forms to trees2. Aligning trees with sentences

2. Identifying frontier nodes3. Extracting minimal rules4. Extracting composed rules

Page 26: An Extended GHKM Algorithm for Inducing  λ -SCFG

Modeling

• Log-linear model + MERT training

• Target

Page 27: An Extended GHKM Algorithm for Inducing  λ -SCFG

Experiments

• Dataset: GEOQUERY– 880 English questions with corresponding Prolog

logical form–

• Metric

Page 28: An Extended GHKM Algorithm for Inducing  λ -SCFG

Experiments

SCFG

PCFG

CCG

Page 29: An Extended GHKM Algorithm for Inducing  λ -SCFG

Experiments

• F-measure for different languages

* en - English, ge - German, el - Greek, th - Thai

Page 30: An Extended GHKM Algorithm for Inducing  λ -SCFG

Experiments

Page 31: An Extended GHKM Algorithm for Inducing  λ -SCFG

Experiments

Page 32: An Extended GHKM Algorithm for Inducing  λ -SCFG