Transcript Document
gUI: Specifying Complete User Interaction Soft computing Laboratory Yonsei University October 25, 2004 Contents • • • • • • • • • 1 Computer human user interfaces – Synthesis Value of empathy in synthesis Computer human user interfaces – Analysis gUI research objectives The role of the anthropomorphic interface metaphor gUI data specification issues Implementation issues Results of the prototype system Conclusion Computer Human User Interfaces – Synthesis • Humane CHIs (Computer Human Interfaces) – Embodied character agents (ECAs), Embodied conversational agents, virtual conversational characters, virtual humans, talking heads • Basic criteria for a standard markup language – Completeness, simplicity, consistency, intuitiveness, abstraction, usability, standardization, evaluation • VHML – An open specification available at the VHML web site – Two applications MetaFace, an ECA The Mentor systems, a dialogue management system 2 Value of Empathy in Synthesis • ECA with emotions or a personality: Increase the feeling of familiarity • Application areas with common anthropomorphic metaphors – Universities – E-commerce – Web guides – Traveling – Interactive games – Knowledge interactive companions 3 Computer Human User Interfaces – Analysis • “Aware”: The user does not know how to perform the task • Gestalt user interfaces (gUIs) – “gestalt”: An organized whole in which each part affects every other part – The mind organizes events and situations as a pattern “big picture” – The entities Similarity, proximity, closure, continuity, membership – Input: A combined “stimulus” of text, button clicks, and analyzed facial, vocal, and emotional gestures – Recognition: Translate the stimulus into the intent 4 gUIs Value of Empathy in Analysis • “more complete” encapsulation of human interaction – The empathic ability of an interface – The extra functionality acquired – An intuitive interface • Overlooked empathic state of the virtual character – Attend: Considering the thoughts and feelings of the user – Engage: Aligning actions, thoughts, and feelings with those of the user – Value: Expressing the value of the user’s interaction – Encourage: Expression encouragement for further interaction – Parting: Suspending dialogue while user performs other tasks – Available: Allowing the user to interrupt for interaction 5 gUI Research Objectives • Formal specification, development, implementation, evaluation, and standardization of a markup language is necessary to provide a stable, consistent base for both industrial use and future research into multi-modal HumanComputer interaction • MPEG-4: Solutions to facial and body animation at the low level • Require a unified standard architecture plus language to control/record the higher level human ECA interaction 6 The Role of the Anthropomorphic Interface Metaphor (1) • Anthropomorphic metaphor – To specify complete user interaction with an ECA – To use the analysis and synthesis of emotion and gesture • Desktop metaphor – The most predominant metaphor used in computer systems – The metaphor’s likeness to our perception of the world • “illusions” – Reduce cognitive overhead when users interact with an anthropomorphic interface – gUI applications become an extension of real life Waste paper basket on the Macintosh desktop The humanity of talking heads 7 The Role of the Anthropomorphic Interface Metaphor (2) • Anthropomorphic web applications – Bryan and Gerchman Browsing: The leisurely navigation along an arbitrary path Searching: Goal oriented finding specific information – A knowledgable ECA (MetaFace) Knows that you are getting frustrated in web searching Offers assistance as a friend • Transcription – Representation, transformation, and storage of analyzed emotions and gestures • Real-time interaction and data representation – Efficient storage and communication of the data – Intelligent television application: handle real-time emotions and gestures to control the function of the TV 8 gUI Data Specification Issues (1) • “Affect recognition” – Proper gestalt understanding of the user requirements – “Non-verbal” behavior and “Non-verbal” communications web site (http://www3.usal.es/~nonverbal/paper.htm) • Temporal window – Differentiate short mixed emotions such as surprise before anger or fear – Detect more lengthy emotions such as gradual happiness as “understanding comes to someone” • Specification of the gUI input data format – At the lowest level: Efficient binary format – At the higher levels: a more structured format aiding in abstraction and cognitive understanding 9 gUI Data Specification Issues (2) 10 gUI Data Specification Issues (3) • Dynamic context transitions • Time-stamping of emotion and gestures: Synchronizing the resulting XML document and providing context • MPEG-4 face and body animation – Efficient communication standard – Emotion and gesture representation for real-time applications • Some problems in VHML data representation – Lack of a consistent set of emotions, and the ability to add more – Lack of ability to show change in different sets of emotions or just a single emotion – Lack of additional attributes for emotion besides intensity 11 gUI Data Specification Issues (4) • A more realistic and extensible semantic structure • Gesture input 12 gUI Data Specification Issues (5) • Issues on specifying gestures – A huge diversity of gestures for the human body to try and identify a base set for all analysis and synthesis applications – Own associated information on each gesture: No standard parameters for all gestures – Culturally dependent gestures • A less naïve format 13 Implementation Issues (1) • MetaFace framework – An example ECA used to test the underlying theories – Based on MPEG-4 and VHML – The ability to vocalize and visually display emotion and gesture – Dialogue management tool language (DMTL) • Evaluate tag: Inference the context 14 Implementation Issues (2) 15 Conclusion • Challenges – The overlap of text input, gesture, and emotion as one coherent stimulus – Standardization of emotion and gesture representation • Complete user interaction with gestalt user interface – Use anthropomorphic metaphors – Reduce the conceptual load – Mentor and MetaFace research project 16