Computing Emotions
- 1. Computing
Emotions
Brian Mac Namee
John Kelleher
Sarah Jane Delany
School of Computing
Dublin Institute of Technology
Applied Intelligence Research Center
- 2. 1 2
Quantitative Recognising
Models of Emotion Emotions
Simulating Interesting
Emotions Questions
3 4
- 4. Lang Model
Arousal
alarmed astonished
angry delighted
annoyed
frustrated
glad
happy
Valence
sad content
satisfied
bored calm
- 10. Emotion Recognition
Emotion recognition has a wide range of uses:
– Call centre monitoring
– Automated tutors
– Music retrieval
– Market research
– Online sentiment analysis
– Law enforcement
- 11. Machine
Learning
Algorithm
Training
Examples
Prediction
Model
- 12. Machine
Learning
Algorithm
Training
Examples
Prediction
Model
Disgust
Happy
Sad
Angry
- 13. Machine
Learning
Algorithm
Training
Examples
Prediction
Model
Positive Negative
- 14. Machine
Learning
Algorithm
Training
Examples
Fear
Prediction
Model
Anger Disgust
Surprise Sadness
Happiness
- 16. Feature Feature Feature Feature Feature
ID 1 2 3 4 5
100 3.6 85 34 22 74
- 17. Feature Feature Feature Feature Feature
ID 1 2 3 4 5
100 3.6 85 34 22 74
Prediction
Model
- 18. Feature Feature Feature Feature Feature
ID 1 2 3 4 5
100 3.6 85 34 22 74
Prediction
Model
Happiness
- 19. Emotion Recognition Examples
Facial Emotion
Recognition
http://www.youtube.com/watch?v=45eLpzk6N34
Sentiment Analysis
http://www.semantic-api.com/demo-statistical-sentiment-analysis.html
- 21. Emotion Simulation
Emotion simulation is required in applications
in which we need to simulate emotions:
– Robotics
– Automated speech systems
– Games/entertainment systems
There are two key parts to emotion simulation:
– A model of emotion generation
– A system for emotion realization
- 22. Lang Model
Arousal
Hand-crafted rules
Insult about how particular
Compliment actions lead to changes
on the dimensions
Valence
within the model
Very application specific
- 23. The OCC Model
The OCC model has a sophisticated logical
formalism for the generation of emotions
Fear equals hope plus anticipated possible
failure
Failure after hope leads to disappointment
- 29. Interesting Questions
What emotions are worth recognising and
simulating and when?
Are there models of human-human interaction
from which lessons about the usefulness of
emotion can be learned?
What are the impacts of better emotional
recognition and simulation for long term human
robot interactions?
Does it matter that the whole thing is a
charade? What happens when users see
behind the curtain?
- 30. Thank You
Brian Mac Namee
John Kelleher
Sarah Jane Delany
School of Computing
Dublin Institute of Technology
Applied Intelligence Research Center