SlideShare a Scribd company logo
1 of 31
Targeting Sentiment
 Supervised Ranking of
Linguistic Configurations

Jason Kessler          Nicolas Nicolov
 Indiana University   J.D. Power and Associates,
                             McGraw Hill
Sentiment Analysis
 ―While the dealership was easy to find and
  the salesman was friendly, the car I bought
  turned out to be a disappointment.‖

• Bag of words:
  – Two positive terms, one negative term
  – Conclusion: author likes the car
What if we knew the sentiment targets?



 ―While the dealership was easy to find and

  the salesman was friendly, the car I

  bought turned out to be a disappointment.‖
Outline
•   Sentiment expressions
•   Finding sentiment targets
•   Previous work
•   Our approach: supervised ranking
•   Evaluation
Sentiment Expressions
• Single or multi-word phrases
  – Express evaluation
• Contextual polarity
  – I like the car (positive)
  – It is a lemon (negative)
  – The camera is not small (negative)
• Assume annotation of sentiment
  expressions, their polarity
Targets
• Target = word or phrase which is the
  object of evaluation
• Sentiment expressions only link to
  physical targets:
        Bill likes to drive.
        Bill likes to drive the car.
• Multiple targets possible:
       — Bill likes the car and the bike.
Targets (2)
Some mentions are not targets.
  – Sue likes1 Al’s car1.
Tricky cases:
  – The car2 frightens2 Mary.
  – Mary4’s dislike3 of Bill’s car3 is a turn-off4 for him.
  – Look at those pancakes5. My mouth is watering5.
Problem
• Given annotation of mentions and
  sentiment expressions
• Identify targets of all sentiment
  expressions
Manual Annotations
                         Entity-level
                                                       sentiment: Positive
John recently purchased a digital camera.
PERSON                                   CAMERA
                   TARGET                    TARGET          TARGET
COREF

  It had a great zoom lens, a mildly disappointingflash,
CAMERA              CAMERA-PART                                  CAMERA-PART
         PART-OF                         PART-OF        Entity-level
                                                        sentiment: Mixed
          TARGET                                                          LESS

  and was very compact.He also considered a Cannon
                               PERSON                            CAMERA
                                        FEATURE-OF
                      TARGET
                                                         MORE

  which, while priced highly        had a better flash.
            CAMERA-FEATURE                               CAMERA-PART

                                                     DIMENSION
Other Annotations
• Sentiment expressions
• Intensifiers, negators, neutralizers,
  committers
• Targets, opinion holders
• Mentions and semantic types
• Coreference, part-of, feature-of, instance-of
• Entity-level sentiment
• Comparisons and their arguments
Corpus Size/Statistics
• Micro-averaged harmonic mean of precision
  between annotator pairs
• Sentiment expressions: 76.84
• Mentions: 87.19
• Targets: 81.55
                                      Sentiment
 Domain    Docs   Tokens Sentences   Expressions   Mentions
   Cars     111   80,560   4,496       3,353        16,953
 Camera      69   38,441   2,218       1,527          9,446
   Total    180 119,001    6,614       4,880        26,399
Baseline - Proximity
• Proximity approach:
  – Nearest mention selected as target
  – Break ties by preferring right-hand mention
  – Breaks on: Sue likes1 Al’s car1.
Baseline – One Hop
   • Run a dependency parser
        – Mentions that govern or are governed by SE
        – Use Stanford dependency parser
        – Partially breaks on:
                                                 DOBJ



                               Sue likes1 Al’s car1.
                                    NSUBJ           POSS




M. de Marneffe, B. MacCartney & C. Manning.
2006. ―Generating typed dependency parses from
phrase structure parses‖. LREC 2006.
Previous Work – Decision List
  • Decision list of dependency paths:
       – Ordered list of 41 labeled dependency paths
         between sentiment expression and mention
       – Top path connecting a sentiment expression to
         a mention     mention is the target
                                                  DOBJ                 DOBJ
…
4. SE –DOBJ Mention
5. SE –NSUBJ Mention          Sue likes1 Al’s car.          It1 upset1 Amy.
…
                                    NSUBJ             POSS     NSUBJ
Sample list slice

Kenneth Bloom, Navendu Garg & Shlomo Argamon. 2007.
―Extracting Appraisal Expressions‖. NAACL-HTL 2007.
Our Approach
• Learning to target from a corpus:
  – Bill likes1 the car1 and Sarah knows it.
  – Classification:
     •   Three independent binary classifier calls
     •   features(like, car) =? Target/Not Target
     •   features(like, Bill) =? Target/Not Target
     •   features(like, Sarah) =? Target/Not Target
Our Approach
• Supervised Ranking
  – Bill likes1 the car1 and Sarah knows it.
  – Rank Bill, car, and Sarah by likelihood of
    being a target of like
     • Ensure car is ranked the highest
  – Learn score function s to appx. rank:
     • Input: features relating sentiment expression,
       mention
     • Output: number that reflects rankings
     • s(features(like, car)) < s(features(like, Bill))
     • s(features(like, car)) < s(features(like, Sarah))
Our Approach
• Learn score function given ranks:
  – Given:
     • My car gets good1 gas milage1.
         – Ranks for good: gas mileage: 0, car: 1, my: 1,
     • It handles2 well2.
         – Ranks for well: handles: 0, it: 1
  – For score function s ensure that:
     • s(features(good, gas mileage)) < s(features(good, car))
     • s(features(good, gas mileage)) < s(features(good, my))
     • s(features(well, handles)) < s(features(well, it))
  – Ensure difference ≥ 1
Our Approach
• Use RankSVM to perform supervised
  ranking
   Joachims, T. 2002. Optimizing search engines using clickthrough data.
   KDD.


• Features
  – Incorporate syntax (dependency parse)
  – Extract labeled-dependency paths between
    mentions and sentiment expressions
Features                                          AUX               DOBJ


                              Paul likes1 to drive the           blue car1
                            NSUBJ                                    DET
                                                        XCOMP

Feature: likes   blue car      Example
# tokens distance               3
# sentiment expressions         0
between
# mentions between              0
Lexical path                    to drive the
Lexical stem path               to drive the
POS path                            TO, VBD, DT                 Encoded as
Stem + labeled dep. path        like :: ↓XCOMP, ↓DOBJ           binary
                                                                features
Labeled dependency path             ↓XCOMP, ↓DOBJ
Semantic type of mention        Car
POS tags of s.exp., mention         VBP, NN
Results – All parts-of-speech
• 10 fold cross validation over all data
90
80
70
60
50                                                   Precision
40                                                   Recall
30                                                   F-score

20
10
 0
     Proximity   One hop   Decision List   RankSVM
Results - Verbs
     Problem:      John likes1 the car1 (-dobj) vs.
                   The car2 upset2 me. (-nsubj)
90
80
70
60
50                                                     Precision
40                                                     Recall
30                                                     F-Score

20
10
0
      Proximity    One hop   Decision List   RankSVM
Results - Adjectives
                                            AMOD
     Problems:    terrible horrible, no good, very bad, movie.
                               DEP
90
80
70
60
50                                                     Precision
40                                                     Recall
30                                                     F-Score

20
10
0
      Proximity   One hop   Decision List    RankSVM
Future work
 – Apply techniques to targeting intensifiers, etc.
 – Inter-sentential targeting
 – Domain adaptation
 – Other approaches Kobayashi et al. (2006), Kim
   and Hovy (2006)

Conclusions
 – Proximity works well
 – Substantial performance gains from supervised
   ranking and syntactic and semantic features
Thank you!

Special thanks to:
• Prof. Martha Palmer
• Prof. Jim Martin
• Dr. Miriam Eckert
• Steliana Ivanova,
• Ron Woodward
• Prof. Michael Gasser
• Jon Elsas
Dependency Features
                                            DOBJ
                      AUX



  Paul    likes1 to         drive     the      blue car1
 NSUBJ                                             AMOD
                                      DET
                        XCOMP

Group sentiment expressions/mentions as single node:

                         AUX                         DOBJ



   Paul     likes1 to         drive     the          blue car1

  NSUBJ                                 DET
                            XCOMP
Dependency Features
                       AUX                 DOBJ




 Paul      likes1 to       drive     the   blue car1

NSUBJ                                DET
                  XCOMP

                                           Like, blue car: ↓XCOMP, ↓DOBJ

                          Great1 car1
                                              Great, car: ↑AMOD
                              AMOD



        ↓ in front of grammatical relation indicates path is followed
        ↑ indicates path is followed in opposite direction
Previous Work
• Kim & Hovy (2006)
    – Use FrameNet-based semantic role labeler on
      sentences with verb/adjective SEs
    – Some frame elements are considered always
      targeting (e.g. stimulus, problem)
                                            stimulus             experiencer


                Bill2’s handling1 of       the situation1 annoyed2 Sam.

                      agent             problem


S.Kim & E.Hovy. 2006. ―Extracting Opinions, Opinion Holders, and Topics Expressed in Online News
Media Text‖. Sentiment and Subjectivity in Text, ACL 2006.
Previous Work
• Kobayashi et al. (2006)
  – Corpus based, statistical machine learning approach
    (Japanese product review corpus)
  – Determining winner reducible to binary classification
      • Bill likes1 the eraser1 and Sarah knows it.
            – Produces training data:
                » Features(Bill, eraser | like, sentence) -> Right
                » Features(eraser, Sarah | like, sentence) -> Left
            – To find like’s target
                » Winner of Bill vs. eraser competes against Sarah
                » Two calls to binary classifier
  – What features to use?, can’t have multiple targets
   Nozomi Kobayashi, Ryu Iida, Kentaro Inui, and Yuji Matsumoto. 2006. Opinion Mining
   on the Web by Extracting Subject-Attribute-Value Relations. In AAAI-CAAW 2006.
Our Approach
  • Supervised ranking (RankSVM):
       – Training data partitioned into subsets
       – Instances xi in each subset (k) are given relative
         rankings, PREF function give difference in ranking
       – Score function s should reflect partial orderings
       – We use SVMLight implementation




Joachims, T. 2002. Optimizing search engines using clickthrough data.
KDD. (Formulation from Lerman et al. EACL’09)
JDPA Sentiment Corpus

More Related Content

More from Jason Kessler

Visualizing Words and Topics with Scattertext
Visualizing Words and Topics with ScattertextVisualizing Words and Topics with Scattertext
Visualizing Words and Topics with ScattertextJason Kessler
 
Natural Language Visualization with Scattertext
Natural Language Visualization with ScattertextNatural Language Visualization with Scattertext
Natural Language Visualization with ScattertextJason Kessler
 
Lexicon Mining for Semiotic Squares: Exploding Binary Classification
Lexicon Mining for Semiotic Squares: Exploding Binary ClassificationLexicon Mining for Semiotic Squares: Exploding Binary Classification
Lexicon Mining for Semiotic Squares: Exploding Binary ClassificationJason Kessler
 
Jason Kessler Problems: What's Wrong with Twitter
Jason Kessler Problems: What's Wrong with TwitterJason Kessler Problems: What's Wrong with Twitter
Jason Kessler Problems: What's Wrong with TwitterJason Kessler
 
Discovering Persuasive Language through Observing Customer Behavior
Discovering Persuasive Language through Observing Customer BehaviorDiscovering Persuasive Language through Observing Customer Behavior
Discovering Persuasive Language through Observing Customer BehaviorJason Kessler
 
Scattertext: A Tool for Visualizing Differences in Language
Scattertext: A Tool for Visualizing Differences in LanguageScattertext: A Tool for Visualizing Differences in Language
Scattertext: A Tool for Visualizing Differences in LanguageJason Kessler
 
From Sentiment to Persuasion Analysis: A Look at Idea Generation Tools
From Sentiment to Persuasion Analysis: A Look at Idea Generation ToolsFrom Sentiment to Persuasion Analysis: A Look at Idea Generation Tools
From Sentiment to Persuasion Analysis: A Look at Idea Generation ToolsJason Kessler
 
The 2010 JDPA Sentiment Corpus for the Automotive Domain
The 2010 JDPA Sentiment Corpus for the Automotive DomainThe 2010 JDPA Sentiment Corpus for the Automotive Domain
The 2010 JDPA Sentiment Corpus for the Automotive DomainJason Kessler
 
Polling the Blogosphere: a Rule-Based Approach to Belief Classification, By J...
Polling the Blogosphere: a Rule-Based Approach to Belief Classification, By J...Polling the Blogosphere: a Rule-Based Approach to Belief Classification, By J...
Polling the Blogosphere: a Rule-Based Approach to Belief Classification, By J...Jason Kessler
 

More from Jason Kessler (9)

Visualizing Words and Topics with Scattertext
Visualizing Words and Topics with ScattertextVisualizing Words and Topics with Scattertext
Visualizing Words and Topics with Scattertext
 
Natural Language Visualization with Scattertext
Natural Language Visualization with ScattertextNatural Language Visualization with Scattertext
Natural Language Visualization with Scattertext
 
Lexicon Mining for Semiotic Squares: Exploding Binary Classification
Lexicon Mining for Semiotic Squares: Exploding Binary ClassificationLexicon Mining for Semiotic Squares: Exploding Binary Classification
Lexicon Mining for Semiotic Squares: Exploding Binary Classification
 
Jason Kessler Problems: What's Wrong with Twitter
Jason Kessler Problems: What's Wrong with TwitterJason Kessler Problems: What's Wrong with Twitter
Jason Kessler Problems: What's Wrong with Twitter
 
Discovering Persuasive Language through Observing Customer Behavior
Discovering Persuasive Language through Observing Customer BehaviorDiscovering Persuasive Language through Observing Customer Behavior
Discovering Persuasive Language through Observing Customer Behavior
 
Scattertext: A Tool for Visualizing Differences in Language
Scattertext: A Tool for Visualizing Differences in LanguageScattertext: A Tool for Visualizing Differences in Language
Scattertext: A Tool for Visualizing Differences in Language
 
From Sentiment to Persuasion Analysis: A Look at Idea Generation Tools
From Sentiment to Persuasion Analysis: A Look at Idea Generation ToolsFrom Sentiment to Persuasion Analysis: A Look at Idea Generation Tools
From Sentiment to Persuasion Analysis: A Look at Idea Generation Tools
 
The 2010 JDPA Sentiment Corpus for the Automotive Domain
The 2010 JDPA Sentiment Corpus for the Automotive DomainThe 2010 JDPA Sentiment Corpus for the Automotive Domain
The 2010 JDPA Sentiment Corpus for the Automotive Domain
 
Polling the Blogosphere: a Rule-Based Approach to Belief Classification, By J...
Polling the Blogosphere: a Rule-Based Approach to Belief Classification, By J...Polling the Blogosphere: a Rule-Based Approach to Belief Classification, By J...
Polling the Blogosphere: a Rule-Based Approach to Belief Classification, By J...
 

Recently uploaded

Understanding the Laravel MVC Architecture
Understanding the Laravel MVC ArchitectureUnderstanding the Laravel MVC Architecture
Understanding the Laravel MVC ArchitecturePixlogix Infotech
 
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 3652toLead Limited
 
"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii Soldatenko"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii SoldatenkoFwdays
 
Streamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project SetupStreamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project SetupFlorian Wilhelm
 
Benefits Of Flutter Compared To Other Frameworks
Benefits Of Flutter Compared To Other FrameworksBenefits Of Flutter Compared To Other Frameworks
Benefits Of Flutter Compared To Other FrameworksSoftradix Technologies
 
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...Patryk Bandurski
 
Scanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL CertsScanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL CertsRizwan Syed
 
Advanced Test Driven-Development @ php[tek] 2024
Advanced Test Driven-Development @ php[tek] 2024Advanced Test Driven-Development @ php[tek] 2024
Advanced Test Driven-Development @ php[tek] 2024Scott Keck-Warren
 
Science&tech:THE INFORMATION AGE STS.pdf
Science&tech:THE INFORMATION AGE STS.pdfScience&tech:THE INFORMATION AGE STS.pdf
Science&tech:THE INFORMATION AGE STS.pdfjimielynbastida
 
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmaticsKotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmaticscarlostorres15106
 
WordPress Websites for Engineers: Elevate Your Brand
WordPress Websites for Engineers: Elevate Your BrandWordPress Websites for Engineers: Elevate Your Brand
WordPress Websites for Engineers: Elevate Your Brandgvaughan
 
Install Stable Diffusion in windows machine
Install Stable Diffusion in windows machineInstall Stable Diffusion in windows machine
Install Stable Diffusion in windows machinePadma Pradeep
 
CloudStudio User manual (basic edition):
CloudStudio User manual (basic edition):CloudStudio User manual (basic edition):
CloudStudio User manual (basic edition):comworks
 
"Federated learning: out of reach no matter how close",Oleksandr Lapshyn
"Federated learning: out of reach no matter how close",Oleksandr Lapshyn"Federated learning: out of reach no matter how close",Oleksandr Lapshyn
"Federated learning: out of reach no matter how close",Oleksandr LapshynFwdays
 
Artificial intelligence in the post-deep learning era
Artificial intelligence in the post-deep learning eraArtificial intelligence in the post-deep learning era
Artificial intelligence in the post-deep learning eraDeakin University
 
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks..."LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...Fwdays
 
Beyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry InnovationBeyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry InnovationSafe Software
 

Recently uploaded (20)

Understanding the Laravel MVC Architecture
Understanding the Laravel MVC ArchitectureUnderstanding the Laravel MVC Architecture
Understanding the Laravel MVC Architecture
 
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
Tech-Forward - Achieving Business Readiness For Copilot in Microsoft 365
 
"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii Soldatenko"Debugging python applications inside k8s environment", Andrii Soldatenko
"Debugging python applications inside k8s environment", Andrii Soldatenko
 
Streamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project SetupStreamlining Python Development: A Guide to a Modern Project Setup
Streamlining Python Development: A Guide to a Modern Project Setup
 
Benefits Of Flutter Compared To Other Frameworks
Benefits Of Flutter Compared To Other FrameworksBenefits Of Flutter Compared To Other Frameworks
Benefits Of Flutter Compared To Other Frameworks
 
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
Integration and Automation in Practice: CI/CD in Mule Integration and Automat...
 
Scanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL CertsScanning the Internet for External Cloud Exposures via SSL Certs
Scanning the Internet for External Cloud Exposures via SSL Certs
 
Advanced Test Driven-Development @ php[tek] 2024
Advanced Test Driven-Development @ php[tek] 2024Advanced Test Driven-Development @ php[tek] 2024
Advanced Test Driven-Development @ php[tek] 2024
 
Science&tech:THE INFORMATION AGE STS.pdf
Science&tech:THE INFORMATION AGE STS.pdfScience&tech:THE INFORMATION AGE STS.pdf
Science&tech:THE INFORMATION AGE STS.pdf
 
Hot Sexy call girls in Panjabi Bagh 🔝 9953056974 🔝 Delhi escort Service
Hot Sexy call girls in Panjabi Bagh 🔝 9953056974 🔝 Delhi escort ServiceHot Sexy call girls in Panjabi Bagh 🔝 9953056974 🔝 Delhi escort Service
Hot Sexy call girls in Panjabi Bagh 🔝 9953056974 🔝 Delhi escort Service
 
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmaticsKotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
Kotlin Multiplatform & Compose Multiplatform - Starter kit for pragmatics
 
WordPress Websites for Engineers: Elevate Your Brand
WordPress Websites for Engineers: Elevate Your BrandWordPress Websites for Engineers: Elevate Your Brand
WordPress Websites for Engineers: Elevate Your Brand
 
Install Stable Diffusion in windows machine
Install Stable Diffusion in windows machineInstall Stable Diffusion in windows machine
Install Stable Diffusion in windows machine
 
CloudStudio User manual (basic edition):
CloudStudio User manual (basic edition):CloudStudio User manual (basic edition):
CloudStudio User manual (basic edition):
 
"Federated learning: out of reach no matter how close",Oleksandr Lapshyn
"Federated learning: out of reach no matter how close",Oleksandr Lapshyn"Federated learning: out of reach no matter how close",Oleksandr Lapshyn
"Federated learning: out of reach no matter how close",Oleksandr Lapshyn
 
Artificial intelligence in the post-deep learning era
Artificial intelligence in the post-deep learning eraArtificial intelligence in the post-deep learning era
Artificial intelligence in the post-deep learning era
 
DMCC Future of Trade Web3 - Special Edition
DMCC Future of Trade Web3 - Special EditionDMCC Future of Trade Web3 - Special Edition
DMCC Future of Trade Web3 - Special Edition
 
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks..."LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...
"LLMs for Python Engineers: Advanced Data Analysis and Semantic Kernel",Oleks...
 
Beyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry InnovationBeyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
Beyond Boundaries: Leveraging No-Code Solutions for Industry Innovation
 
E-Vehicle_Hacking_by_Parul Sharma_null_owasp.pptx
E-Vehicle_Hacking_by_Parul Sharma_null_owasp.pptxE-Vehicle_Hacking_by_Parul Sharma_null_owasp.pptx
E-Vehicle_Hacking_by_Parul Sharma_null_owasp.pptx
 

Targeting Sentiment Expressions through Supervised Ranking of Linguistic Configurations

  • 1. Targeting Sentiment Supervised Ranking of Linguistic Configurations Jason Kessler Nicolas Nicolov Indiana University J.D. Power and Associates, McGraw Hill
  • 2. Sentiment Analysis ―While the dealership was easy to find and the salesman was friendly, the car I bought turned out to be a disappointment.‖ • Bag of words: – Two positive terms, one negative term – Conclusion: author likes the car
  • 3. What if we knew the sentiment targets? ―While the dealership was easy to find and the salesman was friendly, the car I bought turned out to be a disappointment.‖
  • 4. Outline • Sentiment expressions • Finding sentiment targets • Previous work • Our approach: supervised ranking • Evaluation
  • 5. Sentiment Expressions • Single or multi-word phrases – Express evaluation • Contextual polarity – I like the car (positive) – It is a lemon (negative) – The camera is not small (negative) • Assume annotation of sentiment expressions, their polarity
  • 6. Targets • Target = word or phrase which is the object of evaluation • Sentiment expressions only link to physical targets:  Bill likes to drive.  Bill likes to drive the car. • Multiple targets possible: — Bill likes the car and the bike.
  • 7. Targets (2) Some mentions are not targets. – Sue likes1 Al’s car1. Tricky cases: – The car2 frightens2 Mary. – Mary4’s dislike3 of Bill’s car3 is a turn-off4 for him. – Look at those pancakes5. My mouth is watering5.
  • 8. Problem • Given annotation of mentions and sentiment expressions • Identify targets of all sentiment expressions
  • 9. Manual Annotations Entity-level sentiment: Positive John recently purchased a digital camera. PERSON CAMERA TARGET TARGET TARGET COREF It had a great zoom lens, a mildly disappointingflash, CAMERA CAMERA-PART CAMERA-PART PART-OF PART-OF Entity-level sentiment: Mixed TARGET LESS and was very compact.He also considered a Cannon PERSON CAMERA FEATURE-OF TARGET MORE which, while priced highly had a better flash. CAMERA-FEATURE CAMERA-PART DIMENSION
  • 10. Other Annotations • Sentiment expressions • Intensifiers, negators, neutralizers, committers • Targets, opinion holders • Mentions and semantic types • Coreference, part-of, feature-of, instance-of • Entity-level sentiment • Comparisons and their arguments
  • 11.
  • 12. Corpus Size/Statistics • Micro-averaged harmonic mean of precision between annotator pairs • Sentiment expressions: 76.84 • Mentions: 87.19 • Targets: 81.55 Sentiment Domain Docs Tokens Sentences Expressions Mentions Cars 111 80,560 4,496 3,353 16,953 Camera 69 38,441 2,218 1,527 9,446 Total 180 119,001 6,614 4,880 26,399
  • 13. Baseline - Proximity • Proximity approach: – Nearest mention selected as target – Break ties by preferring right-hand mention – Breaks on: Sue likes1 Al’s car1.
  • 14. Baseline – One Hop • Run a dependency parser – Mentions that govern or are governed by SE – Use Stanford dependency parser – Partially breaks on: DOBJ Sue likes1 Al’s car1. NSUBJ POSS M. de Marneffe, B. MacCartney & C. Manning. 2006. ―Generating typed dependency parses from phrase structure parses‖. LREC 2006.
  • 15. Previous Work – Decision List • Decision list of dependency paths: – Ordered list of 41 labeled dependency paths between sentiment expression and mention – Top path connecting a sentiment expression to a mention mention is the target DOBJ DOBJ … 4. SE –DOBJ Mention 5. SE –NSUBJ Mention Sue likes1 Al’s car. It1 upset1 Amy. … NSUBJ POSS NSUBJ Sample list slice Kenneth Bloom, Navendu Garg & Shlomo Argamon. 2007. ―Extracting Appraisal Expressions‖. NAACL-HTL 2007.
  • 16. Our Approach • Learning to target from a corpus: – Bill likes1 the car1 and Sarah knows it. – Classification: • Three independent binary classifier calls • features(like, car) =? Target/Not Target • features(like, Bill) =? Target/Not Target • features(like, Sarah) =? Target/Not Target
  • 17. Our Approach • Supervised Ranking – Bill likes1 the car1 and Sarah knows it. – Rank Bill, car, and Sarah by likelihood of being a target of like • Ensure car is ranked the highest – Learn score function s to appx. rank: • Input: features relating sentiment expression, mention • Output: number that reflects rankings • s(features(like, car)) < s(features(like, Bill)) • s(features(like, car)) < s(features(like, Sarah))
  • 18. Our Approach • Learn score function given ranks: – Given: • My car gets good1 gas milage1. – Ranks for good: gas mileage: 0, car: 1, my: 1, • It handles2 well2. – Ranks for well: handles: 0, it: 1 – For score function s ensure that: • s(features(good, gas mileage)) < s(features(good, car)) • s(features(good, gas mileage)) < s(features(good, my)) • s(features(well, handles)) < s(features(well, it)) – Ensure difference ≥ 1
  • 19. Our Approach • Use RankSVM to perform supervised ranking Joachims, T. 2002. Optimizing search engines using clickthrough data. KDD. • Features – Incorporate syntax (dependency parse) – Extract labeled-dependency paths between mentions and sentiment expressions
  • 20. Features AUX DOBJ Paul likes1 to drive the blue car1 NSUBJ DET XCOMP Feature: likes blue car Example # tokens distance 3 # sentiment expressions 0 between # mentions between 0 Lexical path to drive the Lexical stem path to drive the POS path TO, VBD, DT Encoded as Stem + labeled dep. path like :: ↓XCOMP, ↓DOBJ binary features Labeled dependency path ↓XCOMP, ↓DOBJ Semantic type of mention Car POS tags of s.exp., mention VBP, NN
  • 21. Results – All parts-of-speech • 10 fold cross validation over all data 90 80 70 60 50 Precision 40 Recall 30 F-score 20 10 0 Proximity One hop Decision List RankSVM
  • 22. Results - Verbs Problem: John likes1 the car1 (-dobj) vs. The car2 upset2 me. (-nsubj) 90 80 70 60 50 Precision 40 Recall 30 F-Score 20 10 0 Proximity One hop Decision List RankSVM
  • 23. Results - Adjectives AMOD Problems: terrible horrible, no good, very bad, movie. DEP 90 80 70 60 50 Precision 40 Recall 30 F-Score 20 10 0 Proximity One hop Decision List RankSVM
  • 24. Future work – Apply techniques to targeting intensifiers, etc. – Inter-sentential targeting – Domain adaptation – Other approaches Kobayashi et al. (2006), Kim and Hovy (2006) Conclusions – Proximity works well – Substantial performance gains from supervised ranking and syntactic and semantic features
  • 25. Thank you! Special thanks to: • Prof. Martha Palmer • Prof. Jim Martin • Dr. Miriam Eckert • Steliana Ivanova, • Ron Woodward • Prof. Michael Gasser • Jon Elsas
  • 26. Dependency Features DOBJ AUX Paul likes1 to drive the blue car1 NSUBJ AMOD DET XCOMP Group sentiment expressions/mentions as single node: AUX DOBJ Paul likes1 to drive the blue car1 NSUBJ DET XCOMP
  • 27. Dependency Features AUX DOBJ Paul likes1 to drive the blue car1 NSUBJ DET XCOMP Like, blue car: ↓XCOMP, ↓DOBJ Great1 car1 Great, car: ↑AMOD AMOD ↓ in front of grammatical relation indicates path is followed ↑ indicates path is followed in opposite direction
  • 28. Previous Work • Kim & Hovy (2006) – Use FrameNet-based semantic role labeler on sentences with verb/adjective SEs – Some frame elements are considered always targeting (e.g. stimulus, problem) stimulus experiencer Bill2’s handling1 of the situation1 annoyed2 Sam. agent problem S.Kim & E.Hovy. 2006. ―Extracting Opinions, Opinion Holders, and Topics Expressed in Online News Media Text‖. Sentiment and Subjectivity in Text, ACL 2006.
  • 29. Previous Work • Kobayashi et al. (2006) – Corpus based, statistical machine learning approach (Japanese product review corpus) – Determining winner reducible to binary classification • Bill likes1 the eraser1 and Sarah knows it. – Produces training data: » Features(Bill, eraser | like, sentence) -> Right » Features(eraser, Sarah | like, sentence) -> Left – To find like’s target » Winner of Bill vs. eraser competes against Sarah » Two calls to binary classifier – What features to use?, can’t have multiple targets Nozomi Kobayashi, Ryu Iida, Kentaro Inui, and Yuji Matsumoto. 2006. Opinion Mining on the Web by Extracting Subject-Attribute-Value Relations. In AAAI-CAAW 2006.
  • 30. Our Approach • Supervised ranking (RankSVM): – Training data partitioned into subsets – Instances xi in each subset (k) are given relative rankings, PREF function give difference in ranking – Score function s should reflect partial orderings – We use SVMLight implementation Joachims, T. 2002. Optimizing search engines using clickthrough data. KDD. (Formulation from Lerman et al. EACL’09)