SlideShare a Scribd company logo
1 of 20
STRUCTURED LABELING
TO FACILITATE CONCEPT
EVOLUTION IN MACHINE
LEARNING
Presenter: Hillol Sarker
Authors
Todd Kulesza, Saleema Amershi, Rich
Caruana, Danyel Fisher, Denis Charles
Motivation
 Machine Learning
 We want to train a machine according to some
target concept
 Supervised machine learning needs
consistent labeled data
 e.g., spam filter, email prioritize
 Difficult to obtain
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Problem
 Labeling Consistency is compromised
 Labeler
 Expertise
 Familiarity with concept
 Judgment ability
 Data Contains
 Ambiguity
 Changing distribution
 Concept change over time
Example?Example?
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Semantic Location
Concept EvolutionConcept Evolution
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Existing Approach
 Machine Learning approaches
 Noise-tolerant algorithm
 Multiple labeler
 Majority voting
 Weighting scheme
 Pairwise comparison (A better fit, then B)
 Problem: No human judgment
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Approach
 Conduct series of formative studies
 In order to investigate concept evolution in
practice
 Observations and feedbackfrom these studies
informed final prototype
 Incorporate feedbacks on initial labeler software
 Design a Study
 Evaluate proposed Structured Labeling
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Preliminary Study 1
 Researchers/practitioners create guidelines for
labelers
 Interviewed 2
 Feedbacks
 Guideline creation process is iterative
 Evolves observing new data
 e.g., examples with multiple interpretation
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Preliminary Study 2
 Recruited 11 machine learning expert
 Binary choice task
 Prototype Software
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Preliminary Study 3
 Conducted on 9 of previous 11
participants 4 weekapart
 Using Same Prototype Software
 Same content but shuffled order
Not Significant Difference
Significant Difference
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Incorporate Feedbacks in Study
Software
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Study Software Interface
 Experiment tested 3 interface conditions
 Baseline
 Traditional Mutually Exclusive “Yes”, “No”, “Could be”
 Structured
 Manual Structuring
 Structured Labeling
 Assisted Structuring
 Structured Labeling
+ Automated Assistance
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Study Procedure
 15 participant
 108 items to label
 Fixed task order
 Cooking, travel,
and gardening
 Study Procedure
 Brief Introduction
 Time to practice
 Log interaction in each interface
 Completion of each
task=>Questionnaire
 Completion of 3 task=>Questionnaire
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Result: Group
Group Count
Structured > Baseline (p<0.001)
 Manual > Baseline (p<0.001)
 Assisted > Baseline (p<0.001)
Pages perGroup
Could be < Yes or No
Yes < No
No Could Be Yes
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Result: Revision
Revisited Count
 Manual > Baseline (p<0.005)
 Assisted > Baseline (p<0.005)
Revised Count
Structured > Baseline (p<0.011)
 Manual > Baseline (p<0.006)
 Assisted > Baseline (p<0.024)
First Half Last Half
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Result: Label Quality
 Matric ARI (Adjusted Rand Index)
 Measures Agreement
 Pairs of items that should end up together over all
possible pairs
 Label Quality
 Manual > Baseline (p=0.02)
 Assisted > Baseline (p=0.02)
 Manual ≠ Assisted (P=0.394)
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Result: Labeling
 Labeling Speed
 Manual < Baseline (p=0.003)
 Assisted < Baseline (p<0.001)
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Feedback
 Participant ranked
each tool as their
favorite
 Ho w o fte n did yo ur
concept change?
 Likert-scale
Favorite Lease Favorite
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Summary
 Structured Labeling
 Helps people evolve concept
 Increases label consistency
 at cost of speed
 Can help Machine learning algorithm
 Weight forgroups (e.g., “definitely yes” vs. “yes”)
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Contribution
 Concept evolution causes inconsistent
labeling
 Being first to show its importance
Not Significant Difference Significant Difference
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Critique of work
 Fixed task order used
 e.g., Cooking, travel, and gardening
 Carry over effect
 Limited to supervised learning
 Assisted structuring
 Not always possible
 May bias decision
Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
Thank You

More Related Content

Similar to 2014.chi.structured labeling to facilitate concept evolution in machine learning

Interactive Technologies in Library Instruction: Using Technology and Active ...
Interactive Technologies in Library Instruction: Using Technology and Active ...Interactive Technologies in Library Instruction: Using Technology and Active ...
Interactive Technologies in Library Instruction: Using Technology and Active ...Anthony Holderied
 
Overview and Preliminary Results of Using PolyCAFe for Collaboration Analysis...
Overview and Preliminary Results of Using PolyCAFe for Collaboration Analysis...Overview and Preliminary Results of Using PolyCAFe for Collaboration Analysis...
Overview and Preliminary Results of Using PolyCAFe for Collaboration Analysis...Traian Rebedea
 
Use of online quizzes to support inquiry-based learning in chemical engineering
Use of online quizzes to support inquiry-based learning in chemical engineeringUse of online quizzes to support inquiry-based learning in chemical engineering
Use of online quizzes to support inquiry-based learning in chemical engineeringcilass.slideshare
 
 Instructional Design for the Active: Employing Interactive Technologies and...
	Instructional Design for the Active: Employing Interactive Technologies and...	Instructional Design for the Active: Employing Interactive Technologies and...
 Instructional Design for the Active: Employing Interactive Technologies and...Anthony Holderied
 
Designing Developing Evaluating
Designing Developing EvaluatingDesigning Developing Evaluating
Designing Developing EvaluatingiAttain
 
香港六合彩
香港六合彩香港六合彩
香港六合彩iewsxc
 
Usability Testing Basics: What's it All About? at Web SIG Cleveland
Usability Testing Basics: What's it All About? at Web SIG ClevelandUsability Testing Basics: What's it All About? at Web SIG Cleveland
Usability Testing Basics: What's it All About? at Web SIG ClevelandCarol Smith
 
Ettc Workshop Using A Prs System
Ettc Workshop   Using A Prs SystemEttc Workshop   Using A Prs System
Ettc Workshop Using A Prs Systemheinricb
 
PPT SLIDES
PPT SLIDESPPT SLIDES
PPT SLIDESbutest
 
PPT SLIDES
PPT SLIDESPPT SLIDES
PPT SLIDESbutest
 
Rapid Software Testing: Strategy
Rapid Software Testing: StrategyRapid Software Testing: Strategy
Rapid Software Testing: StrategyTechWell
 
Second presentation o.c.o. technology & paper prototyping
Second presentation  o.c.o. technology & paper prototypingSecond presentation  o.c.o. technology & paper prototyping
Second presentation o.c.o. technology & paper prototypingErika Fisher
 
Web Performance Analysis - TCF Pro 2009
Web Performance Analysis - TCF Pro 2009Web Performance Analysis - TCF Pro 2009
Web Performance Analysis - TCF Pro 2009Guy Ferraiolo
 
7. evalution of interactive system
7. evalution of interactive system7. evalution of interactive system
7. evalution of interactive systemKh Ravy
 
Recommendation System for Design Patterns in Software Development
Recommendation System for Design Patterns in Software DevelopmentRecommendation System for Design Patterns in Software Development
Recommendation System for Design Patterns in Software DevelopmentFrancis Palma
 
Master project - Competitive Co-evolutionary Code-Smells Detection
Master project - Competitive Co-evolutionary Code-Smells DetectionMaster project - Competitive Co-evolutionary Code-Smells Detection
Master project - Competitive Co-evolutionary Code-Smells DetectionMohamed BOUSSAA
 

Similar to 2014.chi.structured labeling to facilitate concept evolution in machine learning (20)

Interactive Technologies in Library Instruction: Using Technology and Active ...
Interactive Technologies in Library Instruction: Using Technology and Active ...Interactive Technologies in Library Instruction: Using Technology and Active ...
Interactive Technologies in Library Instruction: Using Technology and Active ...
 
Overview and Preliminary Results of Using PolyCAFe for Collaboration Analysis...
Overview and Preliminary Results of Using PolyCAFe for Collaboration Analysis...Overview and Preliminary Results of Using PolyCAFe for Collaboration Analysis...
Overview and Preliminary Results of Using PolyCAFe for Collaboration Analysis...
 
Use of online quizzes to support inquiry-based learning in chemical engineering
Use of online quizzes to support inquiry-based learning in chemical engineeringUse of online quizzes to support inquiry-based learning in chemical engineering
Use of online quizzes to support inquiry-based learning in chemical engineering
 
Software Testing
Software TestingSoftware Testing
Software Testing
 
 Instructional Design for the Active: Employing Interactive Technologies and...
	Instructional Design for the Active: Employing Interactive Technologies and...	Instructional Design for the Active: Employing Interactive Technologies and...
 Instructional Design for the Active: Employing Interactive Technologies and...
 
Designing Developing Evaluating
Designing Developing EvaluatingDesigning Developing Evaluating
Designing Developing Evaluating
 
香港六合彩
香港六合彩香港六合彩
香港六合彩
 
Usability Testing Basics: What's it All About? at Web SIG Cleveland
Usability Testing Basics: What's it All About? at Web SIG ClevelandUsability Testing Basics: What's it All About? at Web SIG Cleveland
Usability Testing Basics: What's it All About? at Web SIG Cleveland
 
Exposé Ontology
Exposé OntologyExposé Ontology
Exposé Ontology
 
Ettc Workshop Using A Prs System
Ettc Workshop   Using A Prs SystemEttc Workshop   Using A Prs System
Ettc Workshop Using A Prs System
 
PPT SLIDES
PPT SLIDESPPT SLIDES
PPT SLIDES
 
PPT SLIDES
PPT SLIDESPPT SLIDES
PPT SLIDES
 
Oco tech pres
Oco tech presOco tech pres
Oco tech pres
 
Rapid Software Testing: Strategy
Rapid Software Testing: StrategyRapid Software Testing: Strategy
Rapid Software Testing: Strategy
 
Second presentation o.c.o. technology & paper prototyping
Second presentation  o.c.o. technology & paper prototypingSecond presentation  o.c.o. technology & paper prototyping
Second presentation o.c.o. technology & paper prototyping
 
Web Performance Analysis - TCF Pro 2009
Web Performance Analysis - TCF Pro 2009Web Performance Analysis - TCF Pro 2009
Web Performance Analysis - TCF Pro 2009
 
7. evalution of interactive system
7. evalution of interactive system7. evalution of interactive system
7. evalution of interactive system
 
Rsse12.ppt
Rsse12.pptRsse12.ppt
Rsse12.ppt
 
Recommendation System for Design Patterns in Software Development
Recommendation System for Design Patterns in Software DevelopmentRecommendation System for Design Patterns in Software Development
Recommendation System for Design Patterns in Software Development
 
Master project - Competitive Co-evolutionary Code-Smells Detection
Master project - Competitive Co-evolutionary Code-Smells DetectionMaster project - Competitive Co-evolutionary Code-Smells Detection
Master project - Competitive Co-evolutionary Code-Smells Detection
 

More from BBKuhn

Sound shredding moustafa
Sound shredding moustafaSound shredding moustafa
Sound shredding moustafaBBKuhn
 
Smoking soujanya
Smoking soujanyaSmoking soujanya
Smoking soujanyaBBKuhn
 
Presentation yamin
Presentation yaminPresentation yamin
Presentation yaminBBKuhn
 
Md2k 0219 shang
Md2k 0219 shangMd2k 0219 shang
Md2k 0219 shangBBKuhn
 
Md2 k 04_19_2015
Md2 k 04_19_2015Md2 k 04_19_2015
Md2 k 04_19_2015BBKuhn
 
March19 tun
March19 tunMarch19 tun
March19 tunBBKuhn
 
March12 rahman
March12 rahmanMarch12 rahman
March12 rahmanBBKuhn
 
March12 natarajan
March12 natarajanMarch12 natarajan
March12 natarajanBBKuhn
 
March12 chatterjee
March12 chatterjeeMarch12 chatterjee
March12 chatterjeeBBKuhn
 
March12 alzantot
March12 alzantotMarch12 alzantot
March12 alzantotBBKuhn
 
March5 gao
March5 gaoMarch5 gao
March5 gaoBBKuhn
 
March5 bargar
March5 bargarMarch5 bargar
March5 bargarBBKuhn
 
MD2K Presentation to Stanford Mobilize (1/22/15)
MD2K Presentation to Stanford Mobilize (1/22/15)MD2K Presentation to Stanford Mobilize (1/22/15)
MD2K Presentation to Stanford Mobilize (1/22/15)BBKuhn
 

More from BBKuhn (13)

Sound shredding moustafa
Sound shredding moustafaSound shredding moustafa
Sound shredding moustafa
 
Smoking soujanya
Smoking soujanyaSmoking soujanya
Smoking soujanya
 
Presentation yamin
Presentation yaminPresentation yamin
Presentation yamin
 
Md2k 0219 shang
Md2k 0219 shangMd2k 0219 shang
Md2k 0219 shang
 
Md2 k 04_19_2015
Md2 k 04_19_2015Md2 k 04_19_2015
Md2 k 04_19_2015
 
March19 tun
March19 tunMarch19 tun
March19 tun
 
March12 rahman
March12 rahmanMarch12 rahman
March12 rahman
 
March12 natarajan
March12 natarajanMarch12 natarajan
March12 natarajan
 
March12 chatterjee
March12 chatterjeeMarch12 chatterjee
March12 chatterjee
 
March12 alzantot
March12 alzantotMarch12 alzantot
March12 alzantot
 
March5 gao
March5 gaoMarch5 gao
March5 gao
 
March5 bargar
March5 bargarMarch5 bargar
March5 bargar
 
MD2K Presentation to Stanford Mobilize (1/22/15)
MD2K Presentation to Stanford Mobilize (1/22/15)MD2K Presentation to Stanford Mobilize (1/22/15)
MD2K Presentation to Stanford Mobilize (1/22/15)
 

Recently uploaded

Human & Veterinary Respiratory Physilogy_DR.E.Muralinath_Associate Professor....
Human & Veterinary Respiratory Physilogy_DR.E.Muralinath_Associate Professor....Human & Veterinary Respiratory Physilogy_DR.E.Muralinath_Associate Professor....
Human & Veterinary Respiratory Physilogy_DR.E.Muralinath_Associate Professor....muralinath2
 
THE ROLE OF BIOTECHNOLOGY IN THE ECONOMIC UPLIFT.pptx
THE ROLE OF BIOTECHNOLOGY IN THE ECONOMIC UPLIFT.pptxTHE ROLE OF BIOTECHNOLOGY IN THE ECONOMIC UPLIFT.pptx
THE ROLE OF BIOTECHNOLOGY IN THE ECONOMIC UPLIFT.pptxANSARKHAN96
 
Role of AI in seed science Predictive modelling and Beyond.pptx
Role of AI in seed science  Predictive modelling and  Beyond.pptxRole of AI in seed science  Predictive modelling and  Beyond.pptx
Role of AI in seed science Predictive modelling and Beyond.pptxArvind Kumar
 
TransientOffsetin14CAftertheCarringtonEventRecordedbyPolarTreeRings
TransientOffsetin14CAftertheCarringtonEventRecordedbyPolarTreeRingsTransientOffsetin14CAftertheCarringtonEventRecordedbyPolarTreeRings
TransientOffsetin14CAftertheCarringtonEventRecordedbyPolarTreeRingsSérgio Sacani
 
Phenolics: types, biosynthesis and functions.
Phenolics: types, biosynthesis and functions.Phenolics: types, biosynthesis and functions.
Phenolics: types, biosynthesis and functions.Silpa
 
PSYCHOSOCIAL NEEDS. in nursing II sem pptx
PSYCHOSOCIAL NEEDS. in nursing II sem pptxPSYCHOSOCIAL NEEDS. in nursing II sem pptx
PSYCHOSOCIAL NEEDS. in nursing II sem pptxSuji236384
 
FAIRSpectra - Enabling the FAIRification of Spectroscopy and Spectrometry
FAIRSpectra - Enabling the FAIRification of Spectroscopy and SpectrometryFAIRSpectra - Enabling the FAIRification of Spectroscopy and Spectrometry
FAIRSpectra - Enabling the FAIRification of Spectroscopy and SpectrometryAlex Henderson
 
Reboulia: features, anatomy, morphology etc.
Reboulia: features, anatomy, morphology etc.Reboulia: features, anatomy, morphology etc.
Reboulia: features, anatomy, morphology etc.Silpa
 
300003-World Science Day For Peace And Development.pptx
300003-World Science Day For Peace And Development.pptx300003-World Science Day For Peace And Development.pptx
300003-World Science Day For Peace And Development.pptxryanrooker
 
The Mariana Trench remarkable geological features on Earth.pptx
The Mariana Trench remarkable geological features on Earth.pptxThe Mariana Trench remarkable geological features on Earth.pptx
The Mariana Trench remarkable geological features on Earth.pptxseri bangash
 
module for grade 9 for distance learning
module for grade 9 for distance learningmodule for grade 9 for distance learning
module for grade 9 for distance learninglevieagacer
 
POGONATUM : morphology, anatomy, reproduction etc.
POGONATUM : morphology, anatomy, reproduction etc.POGONATUM : morphology, anatomy, reproduction etc.
POGONATUM : morphology, anatomy, reproduction etc.Silpa
 
Call Girls Ahmedabad +917728919243 call me Independent Escort Service
Call Girls Ahmedabad +917728919243 call me Independent Escort ServiceCall Girls Ahmedabad +917728919243 call me Independent Escort Service
Call Girls Ahmedabad +917728919243 call me Independent Escort Serviceshivanisharma5244
 
Dr. E. Muralinath_ Blood indices_clinical aspects
Dr. E. Muralinath_ Blood indices_clinical  aspectsDr. E. Muralinath_ Blood indices_clinical  aspects
Dr. E. Muralinath_ Blood indices_clinical aspectsmuralinath2
 
Porella : features, morphology, anatomy, reproduction etc.
Porella : features, morphology, anatomy, reproduction etc.Porella : features, morphology, anatomy, reproduction etc.
Porella : features, morphology, anatomy, reproduction etc.Silpa
 
Cyathodium bryophyte: morphology, anatomy, reproduction etc.
Cyathodium bryophyte: morphology, anatomy, reproduction etc.Cyathodium bryophyte: morphology, anatomy, reproduction etc.
Cyathodium bryophyte: morphology, anatomy, reproduction etc.Silpa
 
Atp synthase , Atp synthase complex 1 to 4.
Atp synthase , Atp synthase complex 1 to 4.Atp synthase , Atp synthase complex 1 to 4.
Atp synthase , Atp synthase complex 1 to 4.Silpa
 
Biogenic Sulfur Gases as Biosignatures on Temperate Sub-Neptune Waterworlds
Biogenic Sulfur Gases as Biosignatures on Temperate Sub-Neptune WaterworldsBiogenic Sulfur Gases as Biosignatures on Temperate Sub-Neptune Waterworlds
Biogenic Sulfur Gases as Biosignatures on Temperate Sub-Neptune WaterworldsSérgio Sacani
 
Proteomics: types, protein profiling steps etc.
Proteomics: types, protein profiling steps etc.Proteomics: types, protein profiling steps etc.
Proteomics: types, protein profiling steps etc.Silpa
 

Recently uploaded (20)

Human & Veterinary Respiratory Physilogy_DR.E.Muralinath_Associate Professor....
Human & Veterinary Respiratory Physilogy_DR.E.Muralinath_Associate Professor....Human & Veterinary Respiratory Physilogy_DR.E.Muralinath_Associate Professor....
Human & Veterinary Respiratory Physilogy_DR.E.Muralinath_Associate Professor....
 
THE ROLE OF BIOTECHNOLOGY IN THE ECONOMIC UPLIFT.pptx
THE ROLE OF BIOTECHNOLOGY IN THE ECONOMIC UPLIFT.pptxTHE ROLE OF BIOTECHNOLOGY IN THE ECONOMIC UPLIFT.pptx
THE ROLE OF BIOTECHNOLOGY IN THE ECONOMIC UPLIFT.pptx
 
Role of AI in seed science Predictive modelling and Beyond.pptx
Role of AI in seed science  Predictive modelling and  Beyond.pptxRole of AI in seed science  Predictive modelling and  Beyond.pptx
Role of AI in seed science Predictive modelling and Beyond.pptx
 
TransientOffsetin14CAftertheCarringtonEventRecordedbyPolarTreeRings
TransientOffsetin14CAftertheCarringtonEventRecordedbyPolarTreeRingsTransientOffsetin14CAftertheCarringtonEventRecordedbyPolarTreeRings
TransientOffsetin14CAftertheCarringtonEventRecordedbyPolarTreeRings
 
Phenolics: types, biosynthesis and functions.
Phenolics: types, biosynthesis and functions.Phenolics: types, biosynthesis and functions.
Phenolics: types, biosynthesis and functions.
 
PSYCHOSOCIAL NEEDS. in nursing II sem pptx
PSYCHOSOCIAL NEEDS. in nursing II sem pptxPSYCHOSOCIAL NEEDS. in nursing II sem pptx
PSYCHOSOCIAL NEEDS. in nursing II sem pptx
 
FAIRSpectra - Enabling the FAIRification of Spectroscopy and Spectrometry
FAIRSpectra - Enabling the FAIRification of Spectroscopy and SpectrometryFAIRSpectra - Enabling the FAIRification of Spectroscopy and Spectrometry
FAIRSpectra - Enabling the FAIRification of Spectroscopy and Spectrometry
 
Reboulia: features, anatomy, morphology etc.
Reboulia: features, anatomy, morphology etc.Reboulia: features, anatomy, morphology etc.
Reboulia: features, anatomy, morphology etc.
 
300003-World Science Day For Peace And Development.pptx
300003-World Science Day For Peace And Development.pptx300003-World Science Day For Peace And Development.pptx
300003-World Science Day For Peace And Development.pptx
 
The Mariana Trench remarkable geological features on Earth.pptx
The Mariana Trench remarkable geological features on Earth.pptxThe Mariana Trench remarkable geological features on Earth.pptx
The Mariana Trench remarkable geological features on Earth.pptx
 
module for grade 9 for distance learning
module for grade 9 for distance learningmodule for grade 9 for distance learning
module for grade 9 for distance learning
 
Site Acceptance Test .
Site Acceptance Test                    .Site Acceptance Test                    .
Site Acceptance Test .
 
POGONATUM : morphology, anatomy, reproduction etc.
POGONATUM : morphology, anatomy, reproduction etc.POGONATUM : morphology, anatomy, reproduction etc.
POGONATUM : morphology, anatomy, reproduction etc.
 
Call Girls Ahmedabad +917728919243 call me Independent Escort Service
Call Girls Ahmedabad +917728919243 call me Independent Escort ServiceCall Girls Ahmedabad +917728919243 call me Independent Escort Service
Call Girls Ahmedabad +917728919243 call me Independent Escort Service
 
Dr. E. Muralinath_ Blood indices_clinical aspects
Dr. E. Muralinath_ Blood indices_clinical  aspectsDr. E. Muralinath_ Blood indices_clinical  aspects
Dr. E. Muralinath_ Blood indices_clinical aspects
 
Porella : features, morphology, anatomy, reproduction etc.
Porella : features, morphology, anatomy, reproduction etc.Porella : features, morphology, anatomy, reproduction etc.
Porella : features, morphology, anatomy, reproduction etc.
 
Cyathodium bryophyte: morphology, anatomy, reproduction etc.
Cyathodium bryophyte: morphology, anatomy, reproduction etc.Cyathodium bryophyte: morphology, anatomy, reproduction etc.
Cyathodium bryophyte: morphology, anatomy, reproduction etc.
 
Atp synthase , Atp synthase complex 1 to 4.
Atp synthase , Atp synthase complex 1 to 4.Atp synthase , Atp synthase complex 1 to 4.
Atp synthase , Atp synthase complex 1 to 4.
 
Biogenic Sulfur Gases as Biosignatures on Temperate Sub-Neptune Waterworlds
Biogenic Sulfur Gases as Biosignatures on Temperate Sub-Neptune WaterworldsBiogenic Sulfur Gases as Biosignatures on Temperate Sub-Neptune Waterworlds
Biogenic Sulfur Gases as Biosignatures on Temperate Sub-Neptune Waterworlds
 
Proteomics: types, protein profiling steps etc.
Proteomics: types, protein profiling steps etc.Proteomics: types, protein profiling steps etc.
Proteomics: types, protein profiling steps etc.
 

2014.chi.structured labeling to facilitate concept evolution in machine learning

  • 1. STRUCTURED LABELING TO FACILITATE CONCEPT EVOLUTION IN MACHINE LEARNING Presenter: Hillol Sarker Authors Todd Kulesza, Saleema Amershi, Rich Caruana, Danyel Fisher, Denis Charles
  • 2. Motivation  Machine Learning  We want to train a machine according to some target concept  Supervised machine learning needs consistent labeled data  e.g., spam filter, email prioritize  Difficult to obtain Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 3. Problem  Labeling Consistency is compromised  Labeler  Expertise  Familiarity with concept  Judgment ability  Data Contains  Ambiguity  Changing distribution  Concept change over time Example?Example? Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 4. Semantic Location Concept EvolutionConcept Evolution Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 5. Existing Approach  Machine Learning approaches  Noise-tolerant algorithm  Multiple labeler  Majority voting  Weighting scheme  Pairwise comparison (A better fit, then B)  Problem: No human judgment Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 6. Approach  Conduct series of formative studies  In order to investigate concept evolution in practice  Observations and feedbackfrom these studies informed final prototype  Incorporate feedbacks on initial labeler software  Design a Study  Evaluate proposed Structured Labeling Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 7. Preliminary Study 1  Researchers/practitioners create guidelines for labelers  Interviewed 2  Feedbacks  Guideline creation process is iterative  Evolves observing new data  e.g., examples with multiple interpretation Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 8. Preliminary Study 2  Recruited 11 machine learning expert  Binary choice task  Prototype Software Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 9. Preliminary Study 3  Conducted on 9 of previous 11 participants 4 weekapart  Using Same Prototype Software  Same content but shuffled order Not Significant Difference Significant Difference Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 10. Incorporate Feedbacks in Study Software Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 11. Study Software Interface  Experiment tested 3 interface conditions  Baseline  Traditional Mutually Exclusive “Yes”, “No”, “Could be”  Structured  Manual Structuring  Structured Labeling  Assisted Structuring  Structured Labeling + Automated Assistance Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 12. Study Procedure  15 participant  108 items to label  Fixed task order  Cooking, travel, and gardening  Study Procedure  Brief Introduction  Time to practice  Log interaction in each interface  Completion of each task=>Questionnaire  Completion of 3 task=>Questionnaire Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 13. Result: Group Group Count Structured > Baseline (p<0.001)  Manual > Baseline (p<0.001)  Assisted > Baseline (p<0.001) Pages perGroup Could be < Yes or No Yes < No No Could Be Yes Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 14. Result: Revision Revisited Count  Manual > Baseline (p<0.005)  Assisted > Baseline (p<0.005) Revised Count Structured > Baseline (p<0.011)  Manual > Baseline (p<0.006)  Assisted > Baseline (p<0.024) First Half Last Half Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 15. Result: Label Quality  Matric ARI (Adjusted Rand Index)  Measures Agreement  Pairs of items that should end up together over all possible pairs  Label Quality  Manual > Baseline (p=0.02)  Assisted > Baseline (p=0.02)  Manual ≠ Assisted (P=0.394) Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 16. Result: Labeling  Labeling Speed  Manual < Baseline (p=0.003)  Assisted < Baseline (p<0.001) Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 17. Feedback  Participant ranked each tool as their favorite  Ho w o fte n did yo ur concept change?  Likert-scale Favorite Lease Favorite Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 18. Summary  Structured Labeling  Helps people evolve concept  Increases label consistency  at cost of speed  Can help Machine learning algorithm  Weight forgroups (e.g., “definitely yes” vs. “yes”) Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 19. Contribution  Concept evolution causes inconsistent labeling  Being first to show its importance Not Significant Difference Significant Difference Introduction Preliminary Study Incorporate Feedback Study Result Conclusion
  • 20. Critique of work  Fixed task order used  e.g., Cooking, travel, and gardening  Carry over effect  Limited to supervised learning  Assisted structuring  Not always possible  May bias decision Introduction Preliminary Study Incorporate Feedback Study Result Conclusion Thank You