SlideShare a Scribd company logo
WSD as Distributed Constraint
Optimization Problem

    Author: Siva Reddy, Abhilash Inumella
    Publication: ACL 2010
    Presenter: Po-Han Lin (林伯翰)
Outline

   Background
   WSD as a DCOP
   Experiment
   Conclusion and Future
   References
Outline

   Background
       WSD
       COP
       DCOP
   WSD as a DCOP
   Experiment
   Conclusion and Future
   References
Word Sense Disambiguation

   Word Sense Disambiguation (WSD)
       The problem of WSD can be defined as the task
        of assigning the most appropriate sense to the
        word within a given context.
       WSD is one of the oldest problems in
        computational linguistics which dates back to
        early 1950’s.
       Example:
           Noun orange has at least two senses:
               color or fruit.
Constraint Optimization Problem

   A COP can be defined as a regular constraint
    satisfaction problem in which constraints are
    weighted and the goal is to find a solution
    maximizing the weight of satisfied constraints.



x is a vector residing in a n-dimensional space
f(x) is a scalar valued objective function,
gi(x) = ci for i = 1, …, n and
hj(x) ≦ dj for j = 1, …, m are constraint functions that need to be satisfied.
Distributed Constraint Optimization
Problem
   DCOP is the distributed analogue to
    constraint optimization.
   DCOP is a problem in which a group of
    agents must distributedly choose values for a
    set of variables such that the cost of a set of
    constraints over the variables is either
    minimized or maximized.
       DCOP can be formalized as a tuple
           (A, V, D, C, F)
Distributed Constraint Optimization
Problem - (A, V, D, C, F)

Outline

   Background
   WSD as a DCOP
   Experiment
   Conclusion and Future
   References
WSD as a DCOP


Detial about Constraints
   Part-of-speech (POS 詞性)
       Different POS can decide different domain of words.
           Ex: play has 47 senses, but only 17 senses correspond to noun category.
   Morphology (構詞學)
       Word form (字形)
       Vocabulary (字詞)
           orange has color and fruit two senses, but oranges only be used in the fruit sense.
   Domain information
       Information can be captured using a unary utility function defined for every word.
   Sense Relatedness
       Sense relatedness between senses of two words wi, wj is captured by a function f, where f
        returns sense relatedness (utility) between senses based on sense taxonomy and gloss
        overlaps.
   Discourse
       Many different word, but same sense.
   Collocations
       co-occur word.
       Ex: bank:
           financial institution
           the edge of a river
           if in a given context bank co-occur with money, “financial institution”
Outline

   Background
   WSD as a DCOP
   Experiment
   Conclusion and Future
   References
Experiment

   Experiment Data
       SENSEVAL-2
       SENSEVAL-3
           Ex:
           <instance id="activate.v.bnc.00024693" docsrc="BNC">
               <answer instance="activate.v.bnc.00024693"
                senseid="38201"/>
               <context>
                … which you step on to <head>activate</head> it . Used
                correctly , … .
               </context>
           </instance>
   Answer
       <lexelt item="activate.v">
       <sense id="38201" source="ws"
        wn="activate%2:36:00::" synset="activate actuate
        energize start stimulate" gloss="to initiate action in;
        make active."/>
       <sense id="38202" source="ws"
        wn="activate%2:30:03::" synset="activate" gloss="in
        chemistry, to make more reactive, as by heating."/>
       …
       </lexelt>
Experiment
   Experiment Results
       Dcop:
           This paper method
       Sinha07
           Sinha and Mihalcea, 2007
           page rank algorithm
       Agirre09
           Agirre and Soroa, 2009
           page rank algorithm
           used additional knowledge
               extended WordNet relations
               sense disambiguated gloss.
       MFS
           most frequent sense
           popular back-off heuristic in
            WSD system.
Outline

   Background
   WSD as a DCOP
   Experiment
   Conclusion and Future
   References
Conclusion

   Modelling WSD in a distributed constraint
    optimization framework.
   Showed that this framework is powerful
    enough to encode information from various
    knowledge sources.
FutureWork

   Only used relatedness based utility functions
    derived from WordNet.
   Effect of other knowledge sources remains to
    be evaluated individually and in combination.
   The best possible combination of weights (zk)
    of knowledge sources is yet to be
    engineered.
Outline

   Background
   WSD as a DCOP
   Experiment
   Conclusion and Future
   References
References
   WSD as a Distributed Constraint Optimization Problem
       http://dl.acm.org/citation.cfm?id=1858916
   Distributed Constraint Optimization Problem
       http://www.doc88.com/p-23745051258.html
       http://en.wikipedia.org/wiki/Distributed_constraint_optimization
   Constraint optimization
       http://en.wikipedia.org/wiki/Constraint_optimization
   Word Sense Disambiguation
       http://wordnet.princeton.edu/
       http://203.64.42.21/course/2005/tgbcl/poko/ohcl-13.htm
       http://www.scholarpedia.org/article/Word_sense_disambiguation
   SENSEVAL
       http://www.senseval.org/

More Related Content

PDF
Enhancing Privacy of Confidential Data using K Anonymization
PDF
Ddd ms march 2012 final
PDF
Allerton
PDF
EFFICIENT KNOWLEDGE BASE MANAGEMENT IN DCSP
DOCX
Bt0078, website design
PDF
Simple effective decipherment via combinatorial optimization
PPTX
Word embeddings
PPS
Des Pingouins
Enhancing Privacy of Confidential Data using K Anonymization
Ddd ms march 2012 final
Allerton
EFFICIENT KNOWLEDGE BASE MANAGEMENT IN DCSP
Bt0078, website design
Simple effective decipherment via combinatorial optimization
Word embeddings
Des Pingouins

Similar to Wsd as distributed constraint optimization problem (20)

PPTX
lecture14-distributed-reprennnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnsentations.pptx
PPTX
Tg noh jeju_workshop
PDF
Modeling documents with Generative Adversarial Networks - John Glover
PPTX
Designing, Visualizing and Understanding Deep Neural Networks
PPTX
word vector embeddings in natural languag processing
PPTX
CVDL Unit-5.pptx this is relates to computer vision
PPTX
Data Con LA 2022 - Transformers for NLP
PPT
SNLI_presentation_2
PPTX
Sequence to Sequence Learning with Neural Networks
PPTX
CMSC 723: Computational Linguistics I
PDF
Deep Generative Models
PDF
graduate_thesis (1)
PPTX
The Geometry of Learning
PDF
Paris Lecture 1
PDF
Semantic video classification based on subtitles and domain terminologies
PDF
Mapping Keywords to
PPTX
CNN for modeling sentence
PPTX
Text Mining for Lexicography
PPT
Сергей Кольцов —НИУ ВШЭ —ICBDA 2015
PPTX
WORD SENSE DISAMIGUITION, methods to resolve WSD, wordnet,
lecture14-distributed-reprennnnnnnnnnnnnnnnnnnnnnnnnnnnnnnnsentations.pptx
Tg noh jeju_workshop
Modeling documents with Generative Adversarial Networks - John Glover
Designing, Visualizing and Understanding Deep Neural Networks
word vector embeddings in natural languag processing
CVDL Unit-5.pptx this is relates to computer vision
Data Con LA 2022 - Transformers for NLP
SNLI_presentation_2
Sequence to Sequence Learning with Neural Networks
CMSC 723: Computational Linguistics I
Deep Generative Models
graduate_thesis (1)
The Geometry of Learning
Paris Lecture 1
Semantic video classification based on subtitles and domain terminologies
Mapping Keywords to
CNN for modeling sentence
Text Mining for Lexicography
Сергей Кольцов —НИУ ВШЭ —ICBDA 2015
WORD SENSE DISAMIGUITION, methods to resolve WSD, wordnet,
Ad

Recently uploaded (20)

PPTX
Tartificialntelligence_presentation.pptx
PDF
Hindi spoken digit analysis for native and non-native speakers
PDF
Mushroom cultivation and it's methods.pdf
PDF
August Patch Tuesday
PDF
Hybrid model detection and classification of lung cancer
PPTX
OMC Textile Division Presentation 2021.pptx
PDF
project resource management chapter-09.pdf
PPTX
TechTalks-8-2019-Service-Management-ITIL-Refresh-ITIL-4-Framework-Supports-Ou...
PPTX
KOM of Painting work and Equipment Insulation REV00 update 25-dec.pptx
PPTX
cloud_computing_Infrastucture_as_cloud_p
PDF
1 - Historical Antecedents, Social Consideration.pdf
PDF
ENT215_Completing-a-large-scale-migration-and-modernization-with-AWS.pdf
PDF
A comparative analysis of optical character recognition models for extracting...
PDF
Profit Center Accounting in SAP S/4HANA, S4F28 Col11
PPTX
A Presentation on Artificial Intelligence
PPTX
Programs and apps: productivity, graphics, security and other tools
PDF
Building Integrated photovoltaic BIPV_UPV.pdf
PPTX
SOPHOS-XG Firewall Administrator PPT.pptx
PDF
Heart disease approach using modified random forest and particle swarm optimi...
PDF
Transform Your ITIL® 4 & ITSM Strategy with AI in 2025.pdf
Tartificialntelligence_presentation.pptx
Hindi spoken digit analysis for native and non-native speakers
Mushroom cultivation and it's methods.pdf
August Patch Tuesday
Hybrid model detection and classification of lung cancer
OMC Textile Division Presentation 2021.pptx
project resource management chapter-09.pdf
TechTalks-8-2019-Service-Management-ITIL-Refresh-ITIL-4-Framework-Supports-Ou...
KOM of Painting work and Equipment Insulation REV00 update 25-dec.pptx
cloud_computing_Infrastucture_as_cloud_p
1 - Historical Antecedents, Social Consideration.pdf
ENT215_Completing-a-large-scale-migration-and-modernization-with-AWS.pdf
A comparative analysis of optical character recognition models for extracting...
Profit Center Accounting in SAP S/4HANA, S4F28 Col11
A Presentation on Artificial Intelligence
Programs and apps: productivity, graphics, security and other tools
Building Integrated photovoltaic BIPV_UPV.pdf
SOPHOS-XG Firewall Administrator PPT.pptx
Heart disease approach using modified random forest and particle swarm optimi...
Transform Your ITIL® 4 & ITSM Strategy with AI in 2025.pdf
Ad

Wsd as distributed constraint optimization problem

  • 1. WSD as Distributed Constraint Optimization Problem Author: Siva Reddy, Abhilash Inumella Publication: ACL 2010 Presenter: Po-Han Lin (林伯翰)
  • 2. Outline  Background  WSD as a DCOP  Experiment  Conclusion and Future  References
  • 3. Outline  Background  WSD  COP  DCOP  WSD as a DCOP  Experiment  Conclusion and Future  References
  • 4. Word Sense Disambiguation  Word Sense Disambiguation (WSD)  The problem of WSD can be defined as the task of assigning the most appropriate sense to the word within a given context.  WSD is one of the oldest problems in computational linguistics which dates back to early 1950’s.  Example:  Noun orange has at least two senses:  color or fruit.
  • 5. Constraint Optimization Problem  A COP can be defined as a regular constraint satisfaction problem in which constraints are weighted and the goal is to find a solution maximizing the weight of satisfied constraints. x is a vector residing in a n-dimensional space f(x) is a scalar valued objective function, gi(x) = ci for i = 1, …, n and hj(x) ≦ dj for j = 1, …, m are constraint functions that need to be satisfied.
  • 6. Distributed Constraint Optimization Problem  DCOP is the distributed analogue to constraint optimization.  DCOP is a problem in which a group of agents must distributedly choose values for a set of variables such that the cost of a set of constraints over the variables is either minimized or maximized.  DCOP can be formalized as a tuple  (A, V, D, C, F)
  • 8. Outline  Background  WSD as a DCOP  Experiment  Conclusion and Future  References
  • 9. WSD as a DCOP 
  • 10. Detial about Constraints  Part-of-speech (POS 詞性)  Different POS can decide different domain of words.  Ex: play has 47 senses, but only 17 senses correspond to noun category.  Morphology (構詞學)  Word form (字形)  Vocabulary (字詞)  orange has color and fruit two senses, but oranges only be used in the fruit sense.  Domain information  Information can be captured using a unary utility function defined for every word.  Sense Relatedness  Sense relatedness between senses of two words wi, wj is captured by a function f, where f returns sense relatedness (utility) between senses based on sense taxonomy and gloss overlaps.  Discourse  Many different word, but same sense.  Collocations  co-occur word.  Ex: bank:  financial institution  the edge of a river  if in a given context bank co-occur with money, “financial institution”
  • 11. Outline  Background  WSD as a DCOP  Experiment  Conclusion and Future  References
  • 12. Experiment  Experiment Data  SENSEVAL-2  SENSEVAL-3  Ex:  <instance id="activate.v.bnc.00024693" docsrc="BNC">  <answer instance="activate.v.bnc.00024693" senseid="38201"/>  <context> … which you step on to <head>activate</head> it . Used correctly , … .  </context>  </instance>
  • 13. Answer  <lexelt item="activate.v">  <sense id="38201" source="ws" wn="activate%2:36:00::" synset="activate actuate energize start stimulate" gloss="to initiate action in; make active."/>  <sense id="38202" source="ws" wn="activate%2:30:03::" synset="activate" gloss="in chemistry, to make more reactive, as by heating."/>  …  </lexelt>
  • 14. Experiment  Experiment Results  Dcop:  This paper method  Sinha07  Sinha and Mihalcea, 2007  page rank algorithm  Agirre09  Agirre and Soroa, 2009  page rank algorithm  used additional knowledge  extended WordNet relations  sense disambiguated gloss.  MFS  most frequent sense  popular back-off heuristic in WSD system.
  • 15. Outline  Background  WSD as a DCOP  Experiment  Conclusion and Future  References
  • 16. Conclusion  Modelling WSD in a distributed constraint optimization framework.  Showed that this framework is powerful enough to encode information from various knowledge sources.
  • 17. FutureWork  Only used relatedness based utility functions derived from WordNet.  Effect of other knowledge sources remains to be evaluated individually and in combination.  The best possible combination of weights (zk) of knowledge sources is yet to be engineered.
  • 18. Outline  Background  WSD as a DCOP  Experiment  Conclusion and Future  References
  • 19. References  WSD as a Distributed Constraint Optimization Problem  http://dl.acm.org/citation.cfm?id=1858916  Distributed Constraint Optimization Problem  http://www.doc88.com/p-23745051258.html  http://en.wikipedia.org/wiki/Distributed_constraint_optimization  Constraint optimization  http://en.wikipedia.org/wiki/Constraint_optimization  Word Sense Disambiguation  http://wordnet.princeton.edu/  http://203.64.42.21/course/2005/tgbcl/poko/ohcl-13.htm  http://www.scholarpedia.org/article/Word_sense_disambiguation  SENSEVAL  http://www.senseval.org/