SlideShare a Scribd company logo
1 of 48
Download to read offline
Error-repair Dependency Parsing for
Ungrammatical Texts
ACL 2017
Keisuke Sakaguchi, Matt Post, Benjamin Van Durme
Background & Motivation
Parsing Noisy Text
Learners Essay, Social Media, Speech Transcription
1
©	http://riolindo.org/wp-content/uploads/2013/09/Reading.jpg,	
https://media.licdn.com/mpr/mpr/AAEAAQAAAAAAAA3BAAAAJGFhMDkxODNkLTBlZDQt
NDE2Mi04NjAyLTYyZDM4OTk5ZjA0Yw.jpg,	http://www.trbimg.com/img-
57918d4d/turbine/la-na-pol-donald-trump-convention-speech-transcript-20160721-snap
Background & Motivation
Parsing Noisy Text
Learners Essay, Web Texts, Speech Transcription
2
©	http://riolindo.org/wp-content/uploads/2013/09/Reading.jpg,	
https://media.licdn.com/mpr/mpr/AAEAAQAAAAAAAA3BAAAAJGFhMDkxODNkLTBlZDQt
NDE2Mi04NjAyLTYyZDM4OTk5ZjA0Yw.jpg,	http://www.trbimg.com/img-
57918d4d/turbine/la-na-pol-donald-trump-convention-speech-transcript-20160721-snap
Background & Motivation
Parsing Noisy Text
Learners Essay, Web Texts, Speech Transcription
e.g.,
(incorrect) I look in forward hear from you.
(correct) I look forward to hearing from you.
3
Background & Motivation
Parsing Noisy Text
Learners Essay, Web Texts, Speech Transcription
e.g.,
(incorrect) I look in forward hear from you.
(correct) I look forward to hearing from you.
Error correction
↓
Parsing
Pipeline
4
Background & Motivation
Parsing Noisy Text
Learners Essay, Web Texts, Speech Transcription
e.g.,
(incorrect) I look in forward hear from you.
(correct) I look forward to hearing from you.
Error correction
↓
Parsing
Pipeline
Error-repair
parsing
Joint training
5
Error-repair Dependency Parsing
1. Non-directional Easy-first parsing
(Goldberg and Elhadad, 2010)
2. Three new actions to repair errors
6
Non-directional Easy-first Parsing 7
a brown fox jumped with joy
a brown joywith
joy
fox
a brown
Non-directional Easy-first Parsing 8
a brown fox jumped with joy
a brown joywith
joy
fox
a brown
Pending List
Non-directional Easy-first Parsing
a brown fox jumped with joy
a brown joywith
joy
fox
a brown
9
ATTACHRIGHT(𝑖)
ATTACHLEFT(𝑖)
Iteratively take actions until a complete tree is built.
Non-directional Easy-first Parsing
a brown fox jumped with joy
a brown joywith
joy
fox
a brown
10
Non-directional Easy-first Parsing
ATTACHRIGHT
a brown fox jumped with joy
a brown joywith
joy
fox
a brown
11
Non-directional Easy-first Parsing
a a fox jumped with joy
a brown joywith
joy
fox
a brown
12
Non-directional Easy-first Parsing
ATTACHRIGHT
a a fox jumped with joy
a brown joywith
joy
fox
a brown
13
Non-directional Easy-first Parsing
a brown fox jumped with joy
a brown joywith
joy
fox
a brown
14
Non-directional Easy-first Parsing
ATTACHLEFT
a brown fox jumped with joy
a brown joywith
joy
fox
a brown
15
Non-directional Easy-first Parsing
a brown fox jumped with joy
a brown joywith
joy
fox
a brown
16
Non-directional Easy-first Parsing
ATTACHLEFT
a brown fox jumped with joy
a brown joywith
joy
fox
a brown
17
Non-directional Easy-first Parsing
a brown fox jumped with joy
a brown joywith
joy
fox
a brown
18
Non-directional Easy-first Parsing
ATTACHRIGHT
a brown fox jumped with joy
a brown joywith
joy
fox
a brown
19
Non-directional Easy-first Parsing
a brown fox jumped with joy
a brown joywith
joy
fox
a brown
20
Non-directional Easy-first Parsing
a brown fox jumped with joy
a brown joywith
joy
fox
a brown
21
root
Three new actions to repair errors
SUBSTITUTE (𝑤%)	replaces a token to another
(grammatically more probable) token
DELETE (𝑤%)	removes an unnecessary token
INSERT (𝑤%) inserts a new token at an index i.
22
Three new actions to repair errors
I look in forward xhearx from you
I youyou
23
I look in forward xhearx from you
I youyou
Three new actions to repair errors 24
ATTACHRIGHT
ATTACHLEFT
I look in forward xhearx from you
I youyou
Three new actions to repair errors 25
SUBSTITUTE / DELETE / INSERT
ATTACHRIGHT
I look in forward xhearx from you
I youyou
Three new actions to repair errors 26
I look in forward xhearx from you
I youyou
Three new actions to repair errors 27
ATTACHLEFT
I look in forward xhearx from you
I youyou
Three new actions to repair errors 28
Three new actions to repair errors
I look in forward xhearx from you
I youyou
29
Three new actions to repair errors
SUBSTITUTE
I look in forward xhearx from you
I youyou
30
Three new actions to repair errors
I look in forward hearing from you
I youyou
31
Three new actions to repair errors
DELETE
I look in forward hearing from you
I youyou
32
Three new actions to repair errors
I look forward hearing from from you
I youyou
33
Three new actions to repair errors
INSERT
I look forward hearing from from you
I youyou
34
Three new actions to repair errors
I look forward to hearing from you
I youyou
35
Three new actions to repair errors
ATTACHLEFT
I look forward to hearing from you
I youyou
36
Three new actions to repair errors
I look look to hearing from you
I youyouI forward
37
We are ready to parse noisy texts … ?
Wait!! The new actions may cause infinite loops.
SUB à SUB à SUB à …
INS à DEL à INS à DEL à ...
38
We are ready to parse noisy texts … ?
Wait!! The new actions may cause infinite loops.
SUB à SUB à SUB à …
INS à DEL à INS à DEL à ...
Heuristic constraints to avoid infinite loops
1. Limiting the number of new action operations
2. Substituted token cannot be substituted again
39
Training the parser 40
Model learns which action to take at each time step.
structured perceptron + learning with exploration
(Goldberg and Nivre, 2013)
features: basic linguistic features
(Goldberg and Elhadad 2010)
Training the parser 41
How to know which action is good (i.e., oracle, valid)?
ATTACHLEFT & ATTACHRIGHT (Goldberg and Elhadad, 2010)
1. proposed edge is in the gold parse and
2. the child (to be attached) already has all its children
SUBSTITUTE, DELETE, & INSERT
3. proposed action decreases the (word) edit distance
to the gold (grammatical) sentence.
Experiments
42
Experiment 1 (simulated data)
Dependency parsing on noisy Penn Treebank
Errors injected similarly to Foster and Andersen (2009)
5 most frequent grammatical errors (CoNLL13)
• Determiner (substitution, deletion, insertion)
• Preposition (substitution, deletion, insertion)
• Noun number (singular vs. plural)
• Verb form (tense and aspect)
• Subject verb agreement
Eval: UAS by SParseval (Roark et al., 2006, Favre et al., 2010)
Baseline: pipeline approach (error correction à parsing)
43
44Result (Dependency: UAS)
Experiment 2 (real data)
Grammaticality improvement on real ESL corpus
Treebank of Learner English (Berzak et al., 2016)
Grammaticality score (Heilman et al., 2014)
Regression model with linguistic features
1 (incomprehensible) ~ 4 (perfect)
45
Result (Grammaticality on learner corpus) 46
*
*
Summary
Error-repair Dependency Parsing
1. Non-directional Easy-first Parsing
2. Three new actions to repair errors
Experimental results
1. more robust against grammatical errors
2. improves grammaticality
47
I look in forward xhearx from you
I youyou

More Related Content

Similar to ACL17_Sakaguchi

Tour of language landscape (BuildStuff)
Tour of language landscape (BuildStuff)Tour of language landscape (BuildStuff)
Tour of language landscape (BuildStuff)Yan Cui
 
Lambda The Extreme: Test-Driving a Functional Language
Lambda The Extreme: Test-Driving a Functional LanguageLambda The Extreme: Test-Driving a Functional Language
Lambda The Extreme: Test-Driving a Functional LanguageAccenture | SolutionsIQ
 
Tour of language landscape (code.talks)
Tour of language landscape (code.talks)Tour of language landscape (code.talks)
Tour of language landscape (code.talks)Yan Cui
 
Music recommendations @ MLConf 2014
Music recommendations @ MLConf 2014Music recommendations @ MLConf 2014
Music recommendations @ MLConf 2014Erik Bernhardsson
 
Recent Advances in Natural Language Processing
Recent Advances in Natural Language ProcessingRecent Advances in Natural Language Processing
Recent Advances in Natural Language ProcessingApache MXNet
 
Back To The Future.Key 2
Back To The Future.Key 2Back To The Future.Key 2
Back To The Future.Key 2gueste8cc560
 
graph2tab, a library to convert experimental workflow graphs into tabular for...
graph2tab, a library to convert experimental workflow graphs into tabular for...graph2tab, a library to convert experimental workflow graphs into tabular for...
graph2tab, a library to convert experimental workflow graphs into tabular for...Rothamsted Research, UK
 
Artificial Intelligence for Undergrads
Artificial Intelligence for UndergradsArtificial Intelligence for Undergrads
Artificial Intelligence for UndergradsJose Berengueres
 
Functional Concepts for OOP Developers
Functional Concepts for OOP DevelopersFunctional Concepts for OOP Developers
Functional Concepts for OOP Developersbrweber2
 
Fluent Refactoring (Lone Star Ruby Conf 2013)
Fluent Refactoring (Lone Star Ruby Conf 2013)Fluent Refactoring (Lone Star Ruby Conf 2013)
Fluent Refactoring (Lone Star Ruby Conf 2013)Sam Livingston-Gray
 
The Great Semicolon Debate
The Great Semicolon DebateThe Great Semicolon Debate
The Great Semicolon DebateMike Wilcox
 
Maintainable code
Maintainable codeMaintainable code
Maintainable codeRiverGlide
 
Using Topological Data Analysis on your BigData
Using Topological Data Analysis on your BigDataUsing Topological Data Analysis on your BigData
Using Topological Data Analysis on your BigDataAnalyticsWeek
 
He stopped using for/while loops, you won't believe what happened next!
He stopped using for/while loops, you won't believe what happened next!He stopped using for/while loops, you won't believe what happened next!
He stopped using for/while loops, you won't believe what happened next!François-Guillaume Ribreau
 
Data oriented design and c++
Data oriented design and c++Data oriented design and c++
Data oriented design and c++Mike Acton
 
Learning for sequences - Adam Mathias
Learning for sequences  - Adam MathiasLearning for sequences  - Adam Mathias
Learning for sequences - Adam MathiasDataFest Tbilisi
 
The Last Line Effect
The Last Line EffectThe Last Line Effect
The Last Line EffectAndrey Karpov
 

Similar to ACL17_Sakaguchi (20)

Lf 2021 rates_vii
Lf 2021 rates_viiLf 2021 rates_vii
Lf 2021 rates_vii
 
Tour of language landscape (BuildStuff)
Tour of language landscape (BuildStuff)Tour of language landscape (BuildStuff)
Tour of language landscape (BuildStuff)
 
Lambda The Extreme: Test-Driving a Functional Language
Lambda The Extreme: Test-Driving a Functional LanguageLambda The Extreme: Test-Driving a Functional Language
Lambda The Extreme: Test-Driving a Functional Language
 
Tour of language landscape (code.talks)
Tour of language landscape (code.talks)Tour of language landscape (code.talks)
Tour of language landscape (code.talks)
 
Music recommendations @ MLConf 2014
Music recommendations @ MLConf 2014Music recommendations @ MLConf 2014
Music recommendations @ MLConf 2014
 
Recent Advances in Natural Language Processing
Recent Advances in Natural Language ProcessingRecent Advances in Natural Language Processing
Recent Advances in Natural Language Processing
 
Back To The Future.Key 2
Back To The Future.Key 2Back To The Future.Key 2
Back To The Future.Key 2
 
graph2tab, a library to convert experimental workflow graphs into tabular for...
graph2tab, a library to convert experimental workflow graphs into tabular for...graph2tab, a library to convert experimental workflow graphs into tabular for...
graph2tab, a library to convert experimental workflow graphs into tabular for...
 
Artificial Intelligence for Undergrads
Artificial Intelligence for UndergradsArtificial Intelligence for Undergrads
Artificial Intelligence for Undergrads
 
Functional Concepts for OOP Developers
Functional Concepts for OOP DevelopersFunctional Concepts for OOP Developers
Functional Concepts for OOP Developers
 
Fluent Refactoring (Lone Star Ruby Conf 2013)
Fluent Refactoring (Lone Star Ruby Conf 2013)Fluent Refactoring (Lone Star Ruby Conf 2013)
Fluent Refactoring (Lone Star Ruby Conf 2013)
 
The Great Semicolon Debate
The Great Semicolon DebateThe Great Semicolon Debate
The Great Semicolon Debate
 
Maintainable code
Maintainable codeMaintainable code
Maintainable code
 
Using Topological Data Analysis on your BigData
Using Topological Data Analysis on your BigDataUsing Topological Data Analysis on your BigData
Using Topological Data Analysis on your BigData
 
He stopped using for/while loops, you won't believe what happened next!
He stopped using for/while loops, you won't believe what happened next!He stopped using for/while loops, you won't believe what happened next!
He stopped using for/while loops, you won't believe what happened next!
 
Word2vec and Friends
Word2vec and FriendsWord2vec and Friends
Word2vec and Friends
 
Data oriented design and c++
Data oriented design and c++Data oriented design and c++
Data oriented design and c++
 
Learning for sequences - Adam Mathias
Learning for sequences  - Adam MathiasLearning for sequences  - Adam Mathias
Learning for sequences - Adam Mathias
 
The Last Line Effect
The Last Line EffectThe Last Line Effect
The Last Line Effect
 
Meta learning tutorial
Meta learning tutorialMeta learning tutorial
Meta learning tutorial
 

More from Keisuke Sakaguchi (10)

EMNLP 2021 proScript (summary slides)
EMNLP 2021 proScript (summary slides)EMNLP 2021 proScript (summary slides)
EMNLP 2021 proScript (summary slides)
 
EMNLP 2021 proScript
EMNLP 2021 proScriptEMNLP 2021 proScript
EMNLP 2021 proScript
 
Acl18 sakaguchi
Acl18 sakaguchiAcl18 sakaguchi
Acl18 sakaguchi
 
Ijcnlp17 sakaguchi
Ijcnlp17 sakaguchiIjcnlp17 sakaguchi
Ijcnlp17 sakaguchi
 
TACL16_Sakaguchi
TACL16_SakaguchiTACL16_Sakaguchi
TACL16_Sakaguchi
 
NAACL15_sakaguchi
NAACL15_sakaguchiNAACL15_sakaguchi
NAACL15_sakaguchi
 
BEA12_sakaguchi
BEA12_sakaguchiBEA12_sakaguchi
BEA12_sakaguchi
 
ACL13_sakaguchi
ACL13_sakaguchiACL13_sakaguchi
ACL13_sakaguchi
 
WMT14_sakaguchi
WMT14_sakaguchiWMT14_sakaguchi
WMT14_sakaguchi
 
COLING12_sakaguchi
COLING12_sakaguchiCOLING12_sakaguchi
COLING12_sakaguchi
 

Recently uploaded

Porous Ceramics seminar and technical writing
Porous Ceramics seminar and technical writingPorous Ceramics seminar and technical writing
Porous Ceramics seminar and technical writingrakeshbaidya232001
 
OSVC_Meta-Data based Simulation Automation to overcome Verification Challenge...
OSVC_Meta-Data based Simulation Automation to overcome Verification Challenge...OSVC_Meta-Data based Simulation Automation to overcome Verification Challenge...
OSVC_Meta-Data based Simulation Automation to overcome Verification Challenge...Soham Mondal
 
Software Development Life Cycle By Team Orange (Dept. of Pharmacy)
Software Development Life Cycle By  Team Orange (Dept. of Pharmacy)Software Development Life Cycle By  Team Orange (Dept. of Pharmacy)
Software Development Life Cycle By Team Orange (Dept. of Pharmacy)Suman Mia
 
Introduction and different types of Ethernet.pptx
Introduction and different types of Ethernet.pptxIntroduction and different types of Ethernet.pptx
Introduction and different types of Ethernet.pptxupamatechverse
 
Introduction to IEEE STANDARDS and its different types.pptx
Introduction to IEEE STANDARDS and its different types.pptxIntroduction to IEEE STANDARDS and its different types.pptx
Introduction to IEEE STANDARDS and its different types.pptxupamatechverse
 
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...ranjana rawat
 
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...Dr.Costas Sachpazis
 
the ladakh protest in leh ladakh 2024 sonam wangchuk.pptx
the ladakh protest in leh ladakh 2024 sonam wangchuk.pptxthe ladakh protest in leh ladakh 2024 sonam wangchuk.pptx
the ladakh protest in leh ladakh 2024 sonam wangchuk.pptxhumanexperienceaaa
 
(SHREYA) Chakan Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Esc...
(SHREYA) Chakan Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Esc...(SHREYA) Chakan Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Esc...
(SHREYA) Chakan Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Esc...ranjana rawat
 
HARMONY IN THE NATURE AND EXISTENCE - Unit-IV
HARMONY IN THE NATURE AND EXISTENCE - Unit-IVHARMONY IN THE NATURE AND EXISTENCE - Unit-IV
HARMONY IN THE NATURE AND EXISTENCE - Unit-IVRajaP95
 
Extrusion Processes and Their Limitations
Extrusion Processes and Their LimitationsExtrusion Processes and Their Limitations
Extrusion Processes and Their Limitations120cr0395
 
(RIA) Call Girls Bhosari ( 7001035870 ) HI-Fi Pune Escorts Service
(RIA) Call Girls Bhosari ( 7001035870 ) HI-Fi Pune Escorts Service(RIA) Call Girls Bhosari ( 7001035870 ) HI-Fi Pune Escorts Service
(RIA) Call Girls Bhosari ( 7001035870 ) HI-Fi Pune Escorts Serviceranjana rawat
 
SPICE PARK APR2024 ( 6,793 SPICE Models )
SPICE PARK APR2024 ( 6,793 SPICE Models )SPICE PARK APR2024 ( 6,793 SPICE Models )
SPICE PARK APR2024 ( 6,793 SPICE Models )Tsuyoshi Horigome
 
Model Call Girl in Narela Delhi reach out to us at 🔝8264348440🔝
Model Call Girl in Narela Delhi reach out to us at 🔝8264348440🔝Model Call Girl in Narela Delhi reach out to us at 🔝8264348440🔝
Model Call Girl in Narela Delhi reach out to us at 🔝8264348440🔝soniya singh
 
Microscopic Analysis of Ceramic Materials.pptx
Microscopic Analysis of Ceramic Materials.pptxMicroscopic Analysis of Ceramic Materials.pptx
Microscopic Analysis of Ceramic Materials.pptxpurnimasatapathy1234
 
Processing & Properties of Floor and Wall Tiles.pptx
Processing & Properties of Floor and Wall Tiles.pptxProcessing & Properties of Floor and Wall Tiles.pptx
Processing & Properties of Floor and Wall Tiles.pptxpranjaldaimarysona
 
APPLICATIONS-AC/DC DRIVES-OPERATING CHARACTERISTICS
APPLICATIONS-AC/DC DRIVES-OPERATING CHARACTERISTICSAPPLICATIONS-AC/DC DRIVES-OPERATING CHARACTERISTICS
APPLICATIONS-AC/DC DRIVES-OPERATING CHARACTERISTICSKurinjimalarL3
 
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptxDecoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptxJoão Esperancinha
 

Recently uploaded (20)

Porous Ceramics seminar and technical writing
Porous Ceramics seminar and technical writingPorous Ceramics seminar and technical writing
Porous Ceramics seminar and technical writing
 
OSVC_Meta-Data based Simulation Automation to overcome Verification Challenge...
OSVC_Meta-Data based Simulation Automation to overcome Verification Challenge...OSVC_Meta-Data based Simulation Automation to overcome Verification Challenge...
OSVC_Meta-Data based Simulation Automation to overcome Verification Challenge...
 
Software Development Life Cycle By Team Orange (Dept. of Pharmacy)
Software Development Life Cycle By  Team Orange (Dept. of Pharmacy)Software Development Life Cycle By  Team Orange (Dept. of Pharmacy)
Software Development Life Cycle By Team Orange (Dept. of Pharmacy)
 
Introduction and different types of Ethernet.pptx
Introduction and different types of Ethernet.pptxIntroduction and different types of Ethernet.pptx
Introduction and different types of Ethernet.pptx
 
Introduction to IEEE STANDARDS and its different types.pptx
Introduction to IEEE STANDARDS and its different types.pptxIntroduction to IEEE STANDARDS and its different types.pptx
Introduction to IEEE STANDARDS and its different types.pptx
 
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
(ANVI) Koregaon Park Call Girls Just Call 7001035870 [ Cash on Delivery ] Pun...
 
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
Structural Analysis and Design of Foundations: A Comprehensive Handbook for S...
 
the ladakh protest in leh ladakh 2024 sonam wangchuk.pptx
the ladakh protest in leh ladakh 2024 sonam wangchuk.pptxthe ladakh protest in leh ladakh 2024 sonam wangchuk.pptx
the ladakh protest in leh ladakh 2024 sonam wangchuk.pptx
 
Exploring_Network_Security_with_JA3_by_Rakesh Seal.pptx
Exploring_Network_Security_with_JA3_by_Rakesh Seal.pptxExploring_Network_Security_with_JA3_by_Rakesh Seal.pptx
Exploring_Network_Security_with_JA3_by_Rakesh Seal.pptx
 
(SHREYA) Chakan Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Esc...
(SHREYA) Chakan Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Esc...(SHREYA) Chakan Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Esc...
(SHREYA) Chakan Call Girls Just Call 7001035870 [ Cash on Delivery ] Pune Esc...
 
HARMONY IN THE NATURE AND EXISTENCE - Unit-IV
HARMONY IN THE NATURE AND EXISTENCE - Unit-IVHARMONY IN THE NATURE AND EXISTENCE - Unit-IV
HARMONY IN THE NATURE AND EXISTENCE - Unit-IV
 
Extrusion Processes and Their Limitations
Extrusion Processes and Their LimitationsExtrusion Processes and Their Limitations
Extrusion Processes and Their Limitations
 
(RIA) Call Girls Bhosari ( 7001035870 ) HI-Fi Pune Escorts Service
(RIA) Call Girls Bhosari ( 7001035870 ) HI-Fi Pune Escorts Service(RIA) Call Girls Bhosari ( 7001035870 ) HI-Fi Pune Escorts Service
(RIA) Call Girls Bhosari ( 7001035870 ) HI-Fi Pune Escorts Service
 
SPICE PARK APR2024 ( 6,793 SPICE Models )
SPICE PARK APR2024 ( 6,793 SPICE Models )SPICE PARK APR2024 ( 6,793 SPICE Models )
SPICE PARK APR2024 ( 6,793 SPICE Models )
 
Model Call Girl in Narela Delhi reach out to us at 🔝8264348440🔝
Model Call Girl in Narela Delhi reach out to us at 🔝8264348440🔝Model Call Girl in Narela Delhi reach out to us at 🔝8264348440🔝
Model Call Girl in Narela Delhi reach out to us at 🔝8264348440🔝
 
Microscopic Analysis of Ceramic Materials.pptx
Microscopic Analysis of Ceramic Materials.pptxMicroscopic Analysis of Ceramic Materials.pptx
Microscopic Analysis of Ceramic Materials.pptx
 
9953056974 Call Girls In South Ex, Escorts (Delhi) NCR.pdf
9953056974 Call Girls In South Ex, Escorts (Delhi) NCR.pdf9953056974 Call Girls In South Ex, Escorts (Delhi) NCR.pdf
9953056974 Call Girls In South Ex, Escorts (Delhi) NCR.pdf
 
Processing & Properties of Floor and Wall Tiles.pptx
Processing & Properties of Floor and Wall Tiles.pptxProcessing & Properties of Floor and Wall Tiles.pptx
Processing & Properties of Floor and Wall Tiles.pptx
 
APPLICATIONS-AC/DC DRIVES-OPERATING CHARACTERISTICS
APPLICATIONS-AC/DC DRIVES-OPERATING CHARACTERISTICSAPPLICATIONS-AC/DC DRIVES-OPERATING CHARACTERISTICS
APPLICATIONS-AC/DC DRIVES-OPERATING CHARACTERISTICS
 
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptxDecoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
Decoding Kotlin - Your guide to solving the mysterious in Kotlin.pptx
 

ACL17_Sakaguchi

  • 1. Error-repair Dependency Parsing for Ungrammatical Texts ACL 2017 Keisuke Sakaguchi, Matt Post, Benjamin Van Durme
  • 2. Background & Motivation Parsing Noisy Text Learners Essay, Social Media, Speech Transcription 1 © http://riolindo.org/wp-content/uploads/2013/09/Reading.jpg, https://media.licdn.com/mpr/mpr/AAEAAQAAAAAAAA3BAAAAJGFhMDkxODNkLTBlZDQt NDE2Mi04NjAyLTYyZDM4OTk5ZjA0Yw.jpg, http://www.trbimg.com/img- 57918d4d/turbine/la-na-pol-donald-trump-convention-speech-transcript-20160721-snap
  • 3. Background & Motivation Parsing Noisy Text Learners Essay, Web Texts, Speech Transcription 2 © http://riolindo.org/wp-content/uploads/2013/09/Reading.jpg, https://media.licdn.com/mpr/mpr/AAEAAQAAAAAAAA3BAAAAJGFhMDkxODNkLTBlZDQt NDE2Mi04NjAyLTYyZDM4OTk5ZjA0Yw.jpg, http://www.trbimg.com/img- 57918d4d/turbine/la-na-pol-donald-trump-convention-speech-transcript-20160721-snap
  • 4. Background & Motivation Parsing Noisy Text Learners Essay, Web Texts, Speech Transcription e.g., (incorrect) I look in forward hear from you. (correct) I look forward to hearing from you. 3
  • 5. Background & Motivation Parsing Noisy Text Learners Essay, Web Texts, Speech Transcription e.g., (incorrect) I look in forward hear from you. (correct) I look forward to hearing from you. Error correction ↓ Parsing Pipeline 4
  • 6. Background & Motivation Parsing Noisy Text Learners Essay, Web Texts, Speech Transcription e.g., (incorrect) I look in forward hear from you. (correct) I look forward to hearing from you. Error correction ↓ Parsing Pipeline Error-repair parsing Joint training 5
  • 7. Error-repair Dependency Parsing 1. Non-directional Easy-first parsing (Goldberg and Elhadad, 2010) 2. Three new actions to repair errors 6
  • 8. Non-directional Easy-first Parsing 7 a brown fox jumped with joy a brown joywith joy fox a brown
  • 9. Non-directional Easy-first Parsing 8 a brown fox jumped with joy a brown joywith joy fox a brown Pending List
  • 10. Non-directional Easy-first Parsing a brown fox jumped with joy a brown joywith joy fox a brown 9 ATTACHRIGHT(𝑖) ATTACHLEFT(𝑖) Iteratively take actions until a complete tree is built.
  • 11. Non-directional Easy-first Parsing a brown fox jumped with joy a brown joywith joy fox a brown 10
  • 12. Non-directional Easy-first Parsing ATTACHRIGHT a brown fox jumped with joy a brown joywith joy fox a brown 11
  • 13. Non-directional Easy-first Parsing a a fox jumped with joy a brown joywith joy fox a brown 12
  • 14. Non-directional Easy-first Parsing ATTACHRIGHT a a fox jumped with joy a brown joywith joy fox a brown 13
  • 15. Non-directional Easy-first Parsing a brown fox jumped with joy a brown joywith joy fox a brown 14
  • 16. Non-directional Easy-first Parsing ATTACHLEFT a brown fox jumped with joy a brown joywith joy fox a brown 15
  • 17. Non-directional Easy-first Parsing a brown fox jumped with joy a brown joywith joy fox a brown 16
  • 18. Non-directional Easy-first Parsing ATTACHLEFT a brown fox jumped with joy a brown joywith joy fox a brown 17
  • 19. Non-directional Easy-first Parsing a brown fox jumped with joy a brown joywith joy fox a brown 18
  • 20. Non-directional Easy-first Parsing ATTACHRIGHT a brown fox jumped with joy a brown joywith joy fox a brown 19
  • 21. Non-directional Easy-first Parsing a brown fox jumped with joy a brown joywith joy fox a brown 20
  • 22. Non-directional Easy-first Parsing a brown fox jumped with joy a brown joywith joy fox a brown 21 root
  • 23. Three new actions to repair errors SUBSTITUTE (𝑤%) replaces a token to another (grammatically more probable) token DELETE (𝑤%) removes an unnecessary token INSERT (𝑤%) inserts a new token at an index i. 22
  • 24. Three new actions to repair errors I look in forward xhearx from you I youyou 23
  • 25. I look in forward xhearx from you I youyou Three new actions to repair errors 24 ATTACHRIGHT ATTACHLEFT
  • 26. I look in forward xhearx from you I youyou Three new actions to repair errors 25 SUBSTITUTE / DELETE / INSERT
  • 27. ATTACHRIGHT I look in forward xhearx from you I youyou Three new actions to repair errors 26
  • 28. I look in forward xhearx from you I youyou Three new actions to repair errors 27
  • 29. ATTACHLEFT I look in forward xhearx from you I youyou Three new actions to repair errors 28
  • 30. Three new actions to repair errors I look in forward xhearx from you I youyou 29
  • 31. Three new actions to repair errors SUBSTITUTE I look in forward xhearx from you I youyou 30
  • 32. Three new actions to repair errors I look in forward hearing from you I youyou 31
  • 33. Three new actions to repair errors DELETE I look in forward hearing from you I youyou 32
  • 34. Three new actions to repair errors I look forward hearing from from you I youyou 33
  • 35. Three new actions to repair errors INSERT I look forward hearing from from you I youyou 34
  • 36. Three new actions to repair errors I look forward to hearing from you I youyou 35
  • 37. Three new actions to repair errors ATTACHLEFT I look forward to hearing from you I youyou 36
  • 38. Three new actions to repair errors I look look to hearing from you I youyouI forward 37
  • 39. We are ready to parse noisy texts … ? Wait!! The new actions may cause infinite loops. SUB à SUB à SUB à … INS à DEL à INS à DEL à ... 38
  • 40. We are ready to parse noisy texts … ? Wait!! The new actions may cause infinite loops. SUB à SUB à SUB à … INS à DEL à INS à DEL à ... Heuristic constraints to avoid infinite loops 1. Limiting the number of new action operations 2. Substituted token cannot be substituted again 39
  • 41. Training the parser 40 Model learns which action to take at each time step. structured perceptron + learning with exploration (Goldberg and Nivre, 2013) features: basic linguistic features (Goldberg and Elhadad 2010)
  • 42. Training the parser 41 How to know which action is good (i.e., oracle, valid)? ATTACHLEFT & ATTACHRIGHT (Goldberg and Elhadad, 2010) 1. proposed edge is in the gold parse and 2. the child (to be attached) already has all its children SUBSTITUTE, DELETE, & INSERT 3. proposed action decreases the (word) edit distance to the gold (grammatical) sentence.
  • 44. Experiment 1 (simulated data) Dependency parsing on noisy Penn Treebank Errors injected similarly to Foster and Andersen (2009) 5 most frequent grammatical errors (CoNLL13) • Determiner (substitution, deletion, insertion) • Preposition (substitution, deletion, insertion) • Noun number (singular vs. plural) • Verb form (tense and aspect) • Subject verb agreement Eval: UAS by SParseval (Roark et al., 2006, Favre et al., 2010) Baseline: pipeline approach (error correction à parsing) 43
  • 46. Experiment 2 (real data) Grammaticality improvement on real ESL corpus Treebank of Learner English (Berzak et al., 2016) Grammaticality score (Heilman et al., 2014) Regression model with linguistic features 1 (incomprehensible) ~ 4 (perfect) 45
  • 47. Result (Grammaticality on learner corpus) 46 * *
  • 48. Summary Error-repair Dependency Parsing 1. Non-directional Easy-first Parsing 2. Three new actions to repair errors Experimental results 1. more robust against grammatical errors 2. improves grammaticality 47 I look in forward xhearx from you I youyou