July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo...

25
July 9, 2003 ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University
  • date post

    20-Dec-2015
  • Category

    Documents

  • view

    217
  • download

    1

Transcript of July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo...

Page 1: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003 ACL 20031

An Improved Pattern Model for Automatic IE Pattern Acquisition

Kiyoshi Sudo

Satoshi Sekine

Ralph Grishman

New York University

Page 2: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 20032

Automatic Pattern Acquisition

The cost of manual construction of extraction patterns is very high.

The cost of preparation of annotated data for supervised learning is still high.

The recent trend of the researches on pattern acquisition is un- (semi-) supervised learning.

Page 3: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 20033

Information Extraction

Identifying entities from source text and mapping from source text to pre-defined table.

“A smiling Palestinian suicide bomber triggered a massive explosion in the heavily policed heart of downtown Jerusalem today, …”

Date:Location:Perpetrator:

downtown Jerusalem

A … suicide bomber

today

Page 4: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 20034

Local Context

Local contexts provides a useful information to identify entities.

“A smiling Palestinian suicide bomber triggered a massive explosion in the heavily policed heart of downtown Jerusalem today, …”

Date:Location:Perpetrator:

downtown Jerusalem

A … suicide bomber

today

Page 5: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 20035

Extraction Pattern

Generalize each instance of entity and its local context into an extraction pattern.

“A smiling Palestinian suicide bomber triggered a massive explosion in the heavily policed heart of downtown Jerusalem today, …”

triggered a massive explosion<person>NE category

Association Rule Perpetrator:

Page 6: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 20036

Dependency Tree forPattern Model

Introducing syntax (dependency tree) clarify the relation of arguments with predicates.

triggered

a massive explosionA smiling Palestiniansuicide bomber

heart

heavily policed downtown Jerusalem

today

SBJOBJ

ADV

IN

Page 7: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 20037

Extraction Pattern models

Predicate-Argument model (Yangarber et al. 2000)

– Based on direct relation with a predicate

Chain model (Sudo et al. 2001)

– Based on a chain of modifiers of a predicate

triggered

<person> explosion

triggered

<person>

triggered

heart

downtown Jerusalem

Page 8: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 20038

Predicate-Argument model

Predicate-Argument model is based on the direct relation of a predicate and its arguments.

triggered

a massive explosion<person> heart

heavily policed downtown Jerusalem

<date>

SBJOBJ ADV

IN

Page 9: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 20039

Chain model

Chain model can capture the chain of modifier with an arbitrary depth in the tree, regardless phrasal or clausal boundary.

triggered

a massive explosion<person> heart

heavily policed <location>

<date>

SBJOBJ ADV

IN

(Sudo et al. 2001) reported 5% gain in recall with same level of precision over Predicate-Argument model.

Page 10: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 200310

Problem

Chain model contains only one node at each level of the tree.

triggered

a massive explosion<person> heart

heavily policed downtown Jerusalem

<date>

SBJOBJ ADV

IN

Page 11: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 200311

Problem

Lack of the context can make a pattern too general, causing a false match on irrelevant text.

triggered

a national financial crisisthe Mexican peso last week

SBJOBJ ADV

“The Mexican peso was devalued and triggered a national financial crisis last week.”

Page 12: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 200312

Subtree model

Generalization of Predicate-Argument and Chain model

– Any connected subtree of a

dependency tree will be considered

as a candidate of extraction pattern.

– Give reliable contexts as Predicate-Argument

model does– Capable to capture long-distance relationship

in dependency tree

Page 13: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 200313

Subtree model

Subtree model can provide more relevant contexts, as well as have a flexibility in traversing arbitrary depth in the tree.

triggered

a massive explosion<person> heart

heavily policed downtown Jerusalem

<date>

SBJOBJ ADV

IN

Page 14: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 200314

Experiment

Entity Extraction task– Identify if an NE instance is involved in scenario or not

Management Succession– Person, Organization, Post (Position_Title)

Murder Arrest– Arresting Agency (Organization), Suspect (Person), Charge

– Source: Japanese newspaper 117,109 articles (Mainichi 1995)– Test: accumulated from Mainichi 1994

Succession 148 documents Arrest 205 documents

Page 15: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 200315

Acquisition Method

The target scenario is specified by TREC-like narrative description

– “Management Succession at the level of executives of a company. The topic of interest should not be limited to the promotion inside the company mentioned, but also includes hiring executives from outside the company of their resignation.” [Translated from Japanese]

Preprocessing– Dependency Analysis, NE-tagging

Document RetrievalR

Page 16: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 200316

Acquisition Method

Count all possible subtrees in R– subtree-mining algorithm (Zaki et al. 2002)

– make a Pattern List of those that conform the pattern model Rank each subtree

R

iii df

Ntfscore log

For each subtree i,

number of times subtree ioccurred in the documents in R

Page 17: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 200317

Acquisition Method

Count all possible subtrees in R– subtree-mining algorithm (Zaki et al. 2002)

– make a Pattern List of those that conform the pattern model Rank each subtree

R

iii df

Ntfscore log

For each subtree i,

number of documents inthe source which contain subtree i

Page 18: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 200318

Overlapping patterns

Pattern List contains many overlapping patterns

– (19) (<organization> report)((<organization>-wa) Happyo_suru)

– (480) (<organization> report that … be appointed)((<organization>-wa) (Shunin_suru-to) Happyo-suru)

works as a weight on patterns with more relevant context

[Translated from Japanese]

Page 19: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 200319

comparison

Page 20: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 200320

Unsupervised Parameter Tuning

Unsupervised text classification task by pattern matching

– retrieved … 300 documents retrieved– random … 300 randomly selected

– For each precision-recall curve for , calculate the area that the curve covers.

Pearson correlation coefficient– rp = 0.80 with 2% confidence

Page 21: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 200321

ExtractionPerformance

Page 22: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 200322

Lessons learned

Subtree vs. Chain– Too-general patterns got more penalized for

Subtree model Penalize by Inversed Document Frequency (Subtree, Chain) More scenario-specific patterns got promoted (Subtree)

Page 23: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 200323

Lessons learned

Subtree vs. Predicate-Argument– Patterns with nominalized predicates

Extraction patterns for headlines e.g. (promotion of <post> <person>)

((<person> <post>-no) Shokaku)

– Noun phrase patterns with chain of modifiers e.g. (<post> with ministerial authority)

(((Daihyoken-no (Aru- (<post>)))

[Translated from Japanese]

Page 24: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 200324

Lessons to be learned

Enhanced scoring function by modern IR technique.– Some techniques directly helps pattern acquisition

e.g. relevance feedback– However, note the crucial difference between Pattern

acquisition and IR Same pattern does not appear twice in a document.

Generic variable instead of sticking to Named Entity categories as place holder.

– How robust can a pattern be without semantic restriction?

Page 25: July 9, 2003ACL 2003 1 An Improved Pattern Model for Automatic IE Pattern Acquisition Kiyoshi Sudo Satoshi Sekine Ralph Grishman New York University.

July 9, 2003ACL 200325

Conclusion

We proposed Subtree model as a generalization of– Predicate-Argument model– Chain model

Subtree model patterns overly performed better than other models in Entity Extraction tasks.

Scoring function needs a special consideration for overlapping patterns.

Unsupervised parameter tuning by text classification task.