際際滷

際際滷Share a Scribd company logo

Wordnet-Affect [IIT-Bombay]

Sagar Ahire
Sagar Ahire

The document discusses Wordnet-Affect, an extension of Wordnet that adds affective information. It was built by manually adding affect labels and information to 1903 terms, then projecting this information to Wordnet synsets. The extension was done semi-automatically using Wordnet relations. Possible applications of Wordnet-Affect include sentiment analysis, verbal expressiveness for conversational agents, and computer-assisted creativity. In general, affective computing aims to enable computers to perceive and express emotion and can be applied in areas like adaptive entertainment and modifying environments.

1 of 33
Downloaded 48 times
Paper Presentation Wordnet Affect by Carlo Strapparava, Alessandro Valitutti Sagar Ahire [133050073]
Roadmap Introduction to Affective Computing Introduction to Wordnet-Affect Process of Building Wordnet-Affect Possible Applications
Roadmap: We Are Here Introduction to Affective Computing Introduction to Wordnet-Affect Process of Building Wordnet-Affect Possible Applications
Affective Computing Refers to processing of emotions Enables computers to perceive and/or express affect Does NOT refer to: Creating emotional computers Introducing unreliability/unpredictability in computers
Need of Affective Computing Importance of emotion to humans Vast wealth of information conveyed via emotions [Sentic Modulation] Body language Facial expressions Voice cues Ability to enhance communication with computers
Need of Affective Computing Importance of emotion to humans Vast wealth of information conveyed via emotions [Sentic Modulation] Body language Facial expressions Voice cues Ability to enhance communication with computers
Need of Affective Computing Importance of emotion to humans Vast wealth of information conveyed via emotions [Sentic Modulation] Body language Facial expressions Voice cues Ability to enhance communication with computers
Types of Affective Computers Cannot Express Can Express Cannot Perceive Most computers today Use of smileys, etc Can Perceive Teaching computers Maximally user-friendly* * Does NOT imply being driven by emotion blue = Open Research Areas
Roadmap: We Are Here Introduction to Affective Computing Introduction to Wordnet-Affect Process of Building Wordnet-Affect Possible Applications
Wordnet-Affect Linguistic resource for affect Wordnet-Affect = Wordnet + Affect Information
Need of Wordnet-Affect Haphazard process of selecting emotion words Lack of emotion features Need of distinguishing different senses of the same word
Affect Information A-label Joy [joy, elation] Love [love, worship] Apprehension [trepidation, foreboding] Sadness [misery, sorrow] Surprise [surprise, astonishment] Apathy [apathy, emotionlessness], etc. POSR: Different POS but same emotion [cheer, cheerful, cheerfulness, cheerfully]
Affect Information Valence Positive Negative Ambiguous Neutral
Partial Graph of WN-Affect root mental-state physical-state cognitive-state joy love worship situation ... affective-state mood positive-emotion behaviour emotion negative-emotion ... sadness neutral-emotion shame ... apathy ambiguousemotion pensiveness
Roadmap: We Are Here Introduction to Affective Computing Introduction to Wordnet-Affect Process of Building Wordnet-Affect Possible Applications
Stages in Building Wordnet-Affect Core [Manual] Extension [Semi-Automatic]
Building Wordnet-Affect: Core Manual resource called Affect created with 1903 terms (not synsets) Each term was given associated affect information Information was then projected from the terms to Wordnet synsets
Building Wordnet-Affect: Extension Extension was performed using the following Wordnet relations (similar to Sentiwordnet): Direct antonymy Similarity Derived from Pertains to Attribute Also see
Building Wordnet-Affect: Extension For other relations, similar expansion was performed, but the synsets had to be filtered manually. Hypernym of cheer is attribute Hyponym of murder is tyrannicide
Roadmap: We Are Here Introduction to Affective Computing Introduction to Wordnet-Affect Process of Building Wordnet-Affect Possible Applications
Possible Applications Wordnet-Affect in particular Affect Computing in general
Applications of WordnetAffect Computer-Assisted Creativity Automated Personalized Advertisements Verbal Expressivity for Conversational Agents Spoken Tutorials Sentiment Analysis Affective Text Sensing
Applications of WordnetAffect Computer-Assisted Creativity Automated Personalized Advertisements Verbal Expressivity for Conversational Agents Spoken Tutorials Sentiment Analysis Affective Text Sensing
Applications of WordnetAffect Computer-Assisted Creativity Automated Personalized Advertisements Verbal Expressivity for Conversational Agents Spoken Tutorials Sentiment Analysis Affective Text Sensing
Applications of Affective Computing Adaptive Entertainment Automated DJing / Music concerts Expressive E-mail Avoiding misunderstandings over the internet by encoding affect in written text Film Processing & Viewing Skip ahead to the interesting part Environment Modifications Temperature, Pressure, Color, etc Aesthetic Entertainment Capturing the subjectivity of art
Applications of Affective Computing Adaptive Entertainment Automated DJing / Music concerts Expressive E-mail Avoiding misunderstandings over the internet by encoding affect in written text Film Processing & Viewing Skip ahead to the interesting part Environment Modifications Temperature, Pressure, Color, etc Aesthetic Entertainment Capturing the subjectivity of art
Applications of Affective Computing Adaptive Entertainment Automated DJing / Music concerts Expressive E-mail Avoiding misunderstandings over the internet by encoding affect in written text Film Processing & Viewing Skip ahead to the interesting part Environment Modifications Temperature, Pressure, Color, etc Aesthetic Entertainment Capturing the subjectivity of art
Applications of Affective Computing Adaptive Entertainment Automated DJing / Music concerts Expressive E-mail Avoiding misunderstandings over the internet by encoding affect in written text Film Processing & Viewing Skip ahead to the interesting part Environment Modifications Temperature, Pressure, Color, etc Aesthetic Entertainment Capturing the subjectivity of art
Applications of Affective Computing Adaptive Entertainment Automated DJing / Music concerts Expressive E-mail Avoiding misunderstandings over the internet by encoding affect in written text Film Processing & Viewing Skip ahead to the interesting part Environment Modifications Temperature, Pressure, Color, etc Aesthetic Entertainment Capturing the subjectivity of art
Applications of Affective Computing Knowledge Implications Emotion Theory Philosophy Mental Health Physical Health
Major References WordNet-Affect: An Affective Extension of WordNet by Carlo Strapparava and Alessandro Valitutti, 2004
Other References Affective Computing by Picard, 1995; see also I, Robot by Isaac Asimov 2001: A Space Odyssey by Arthur Clarke / Stanley Kubrick The Affective Weight of Lexicon by Carlo Strapparava, Alessandro Valitutti, Oliviero Stock, 2006
Further Plan Lexicon-based Methods in Sentiment Analysis by Maite Taboada, Julian Brooke, Milan Tofiloski, Kimberly Voll, Manfred Stede, 2011 - DONE Automatic Generation of Lexical Resources for Opinion Mining by Andrea Esuli (PhD Thesis), 2008

More Related Content

Wordnet-Affect [IIT-Bombay]

  • 1. Paper Presentation Wordnet Affect by Carlo Strapparava, Alessandro Valitutti Sagar Ahire [133050073]
  • 2. Roadmap Introduction to Affective Computing Introduction to Wordnet-Affect Process of Building Wordnet-Affect Possible Applications
  • 3. Roadmap: We Are Here Introduction to Affective Computing Introduction to Wordnet-Affect Process of Building Wordnet-Affect Possible Applications
  • 4. Affective Computing Refers to processing of emotions Enables computers to perceive and/or express affect Does NOT refer to: Creating emotional computers Introducing unreliability/unpredictability in computers
  • 5. Need of Affective Computing Importance of emotion to humans Vast wealth of information conveyed via emotions [Sentic Modulation] Body language Facial expressions Voice cues Ability to enhance communication with computers
  • 6. Need of Affective Computing Importance of emotion to humans Vast wealth of information conveyed via emotions [Sentic Modulation] Body language Facial expressions Voice cues Ability to enhance communication with computers
  • 7. Need of Affective Computing Importance of emotion to humans Vast wealth of information conveyed via emotions [Sentic Modulation] Body language Facial expressions Voice cues Ability to enhance communication with computers
  • 8. Types of Affective Computers Cannot Express Can Express Cannot Perceive Most computers today Use of smileys, etc Can Perceive Teaching computers Maximally user-friendly* * Does NOT imply being driven by emotion blue = Open Research Areas
  • 9. Roadmap: We Are Here Introduction to Affective Computing Introduction to Wordnet-Affect Process of Building Wordnet-Affect Possible Applications
  • 10. Wordnet-Affect Linguistic resource for affect Wordnet-Affect = Wordnet + Affect Information
  • 11. Need of Wordnet-Affect Haphazard process of selecting emotion words Lack of emotion features Need of distinguishing different senses of the same word
  • 12. Affect Information A-label Joy [joy, elation] Love [love, worship] Apprehension [trepidation, foreboding] Sadness [misery, sorrow] Surprise [surprise, astonishment] Apathy [apathy, emotionlessness], etc. POSR: Different POS but same emotion [cheer, cheerful, cheerfulness, cheerfully]
  • 14. Partial Graph of WN-Affect root mental-state physical-state cognitive-state joy love worship situation ... affective-state mood positive-emotion behaviour emotion negative-emotion ... sadness neutral-emotion shame ... apathy ambiguousemotion pensiveness
  • 15. Roadmap: We Are Here Introduction to Affective Computing Introduction to Wordnet-Affect Process of Building Wordnet-Affect Possible Applications
  • 16. Stages in Building Wordnet-Affect Core [Manual] Extension [Semi-Automatic]
  • 17. Building Wordnet-Affect: Core Manual resource called Affect created with 1903 terms (not synsets) Each term was given associated affect information Information was then projected from the terms to Wordnet synsets
  • 18. Building Wordnet-Affect: Extension Extension was performed using the following Wordnet relations (similar to Sentiwordnet): Direct antonymy Similarity Derived from Pertains to Attribute Also see
  • 19. Building Wordnet-Affect: Extension For other relations, similar expansion was performed, but the synsets had to be filtered manually. Hypernym of cheer is attribute Hyponym of murder is tyrannicide
  • 20. Roadmap: We Are Here Introduction to Affective Computing Introduction to Wordnet-Affect Process of Building Wordnet-Affect Possible Applications
  • 21. Possible Applications Wordnet-Affect in particular Affect Computing in general
  • 22. Applications of WordnetAffect Computer-Assisted Creativity Automated Personalized Advertisements Verbal Expressivity for Conversational Agents Spoken Tutorials Sentiment Analysis Affective Text Sensing
  • 23. Applications of WordnetAffect Computer-Assisted Creativity Automated Personalized Advertisements Verbal Expressivity for Conversational Agents Spoken Tutorials Sentiment Analysis Affective Text Sensing
  • 24. Applications of WordnetAffect Computer-Assisted Creativity Automated Personalized Advertisements Verbal Expressivity for Conversational Agents Spoken Tutorials Sentiment Analysis Affective Text Sensing
  • 25. Applications of Affective Computing Adaptive Entertainment Automated DJing / Music concerts Expressive E-mail Avoiding misunderstandings over the internet by encoding affect in written text Film Processing & Viewing Skip ahead to the interesting part Environment Modifications Temperature, Pressure, Color, etc Aesthetic Entertainment Capturing the subjectivity of art
  • 26. Applications of Affective Computing Adaptive Entertainment Automated DJing / Music concerts Expressive E-mail Avoiding misunderstandings over the internet by encoding affect in written text Film Processing & Viewing Skip ahead to the interesting part Environment Modifications Temperature, Pressure, Color, etc Aesthetic Entertainment Capturing the subjectivity of art
  • 27. Applications of Affective Computing Adaptive Entertainment Automated DJing / Music concerts Expressive E-mail Avoiding misunderstandings over the internet by encoding affect in written text Film Processing & Viewing Skip ahead to the interesting part Environment Modifications Temperature, Pressure, Color, etc Aesthetic Entertainment Capturing the subjectivity of art
  • 28. Applications of Affective Computing Adaptive Entertainment Automated DJing / Music concerts Expressive E-mail Avoiding misunderstandings over the internet by encoding affect in written text Film Processing & Viewing Skip ahead to the interesting part Environment Modifications Temperature, Pressure, Color, etc Aesthetic Entertainment Capturing the subjectivity of art
  • 29. Applications of Affective Computing Adaptive Entertainment Automated DJing / Music concerts Expressive E-mail Avoiding misunderstandings over the internet by encoding affect in written text Film Processing & Viewing Skip ahead to the interesting part Environment Modifications Temperature, Pressure, Color, etc Aesthetic Entertainment Capturing the subjectivity of art
  • 30. Applications of Affective Computing Knowledge Implications Emotion Theory Philosophy Mental Health Physical Health
  • 31. Major References WordNet-Affect: An Affective Extension of WordNet by Carlo Strapparava and Alessandro Valitutti, 2004
  • 32. Other References Affective Computing by Picard, 1995; see also I, Robot by Isaac Asimov 2001: A Space Odyssey by Arthur Clarke / Stanley Kubrick The Affective Weight of Lexicon by Carlo Strapparava, Alessandro Valitutti, Oliviero Stock, 2006
  • 33. Further Plan Lexicon-based Methods in Sentiment Analysis by Maite Taboada, Julian Brooke, Milan Tofiloski, Kimberly Voll, Manfred Stede, 2011 - DONE Automatic Generation of Lexical Resources for Opinion Mining by Andrea Esuli (PhD Thesis), 2008