Confetti
Master's Project for summer term 2009 and winter term 2009/2010
Lead by
Prof. Dr. Rainer Malaka, University of Bremen
Robert Porzel, University of Bremen
AbstractEven after 50 years the program Eliza, designed by Joseph Weizenbaum in 1966, remains popular and the vision of an engaging conversational interface is still a topical one. Such a conversational interface, for example, can be realized as an interactive character that can produce and understand human language as well as gestures, facial expressions and other modes of expression (see Figure 1). While Weizenbaum’s Eliza parodied a therapist, and worked largely by rephrasing many of the user’s statements as questions and posing them back to the user, today’s multimodal interfaces aim to provide natural communication and interaction with information technology in many domains outside psychiatry.
The aim of the project is to apply state of the art technology from the fields of intelligent user interfaces, adaptive media and natural language and image processing to test to what degree human conversational capabilities can be implemented today. Therefore several subsequent questions emerge that can become one of the foci of the master project:
- What types of communication are needed for natural interaction? Does human language alone suffice or do we also need non-verbal communication, emotions and more. ?
- What kind of dialog design is suitable and how can we manage the interaction, e.g. with dialog signals (“ahm”, “err”, “hmm”), gestures, nods and the like?
- How much shared attention and shared world and contextual knowledge is necessary to communicate successfully?
- How can we measure communicative success, engagement and conversational capabilities?
- How can the system extend its capabilities and learn to understand the user better and to react in more felicitous ways?
The research areas relevant for this project can be:
- Human computer interaction: Interfaces mediate the action between the user and the computer. The idea behind multimodal interaction is the concept involves taking advantage of the full range of human senses and motor skills. Such an interface has to transform digital data into humanly perceivable forms and vice versa.
- Active and adaptive media: Interactive media must be able to react adequately to complex situations and have to adapt their behaviour to their context. In order to achieve natural and engaging interaction one has to consider going beyond simple stimulus-response systems and enable the system to acquire relevant conceptual and contextual knowledge. Additionally, they could reflect on their situation and to actively communicate their state to the user in an intelligible way.
- Machine learning: Many components involved in multimodal processing require a variety statistical classification techniques and both online- and offline-learning are possible.
Through the use of algorithms and tools from machine learning, active and adaptive media and human computer interaction, e.g. language technology, a new dimension in engaging and adaptive interfaces will, therefore, be explored in the Master Project. The group “Digital Media” at the University of Bremen, directed by Rainer Malaka, has a long-term record of research in these areas. The project will have access to a number of active research projects and a the graduate school “Advances in Digital Media”.
Learning Objectives
- Learning about methods for creating innovative user interfaces
- Learning about machine learning approaches
- Learning about evaluation methods
- Working in interdisciplinary, multi-cultural software development teams, learning about project management, cooperative work and participatory design
Intended Outcomes
- Usability Experiments
- Prototype for human computer interaction
- Documentation
PrerequisitesInterested students should have a high interest in both design and development of complex digital media Applications as well as:
- Practical experience in programming and interface design
- Knowledge about different approaches for software development and project management
- Background in artificial intelligence and digital media
Mandatory preparing lectures
- Mastering Digital Media / Requirements Engineering at the University of Bremen
- Machine Learning at the University of Bremen
Selected Readings
- Tom Mitchell (1997) Machine Learning. McGraw-Hill
- Daniel Jurafsky and James H. Martin (2000) Speech and
Language Processing. Prentice Hall
- Jerome A. Feldman (2006) From Molecule to
Metaphor. A Neural Theory of Language. MIT Press.
- Luc Steels (1999) The Talking Heads Experiment
- Paul Dourish (2001) Where The Action Is: The Foundations
of Embodied Interaction, MIT Press October 2001
ContactDigital Media
Prof. Dr. Rainer Malaka
Robert Porzel
Digital Media
Bibliothekstr. 1
D-28359 Bremen
T: ++49 (0) 421 64400
F: ++49 (0) 421 218 8751
E:
Diese E-Mail Adresse ist gegen Spam-Bots geschützt, Sie müssen Javascript aktivieren, damit Sie es sehen können
http://medien.informatik.uni-bremen.de/
Abstract as PDF
|
|
Letzte Aktualisierung ( Montag, 24 März 2008 )
|
Master Studies 
