![]() |
Language, Learning and Multimodal Systems: Technological, Cognitive and Philosophical Perspectives
Timo Honkela Helsinki University of Technology Adaptive Informatics Research Center Date: Friday, December 1,
2006 The efforts to create systems that understand natural language expressions, translate from one language to another, or respond intelligently to spoken questions and commands have had only limited success. In this presentation, I will discuss the limitations of the traditional approaches that are based on certain assumptions on the nature of language and communication. An important starting point here is the inherent subjectivity and multimodality of human experience. An alternative approach is based on technologies that make the systems learn to understand. This can be achieved, for instance, by applying statistical machine learning methods or artificial neural network models. Moreover, one can model communities of adaptive agents and the rich interactions between them rather than an individual intelligent system.
I will present a number of practical results based on the use of methods such the self-organizing map and independent component analysis. These are unsupervised learning methods that give raise to (semi)autonomous representation of rich linguistic and multimodal data. The examples discussed in more detail include:
- The Pockets Full of Memories interactive museum installation - Emergence of ontological relations based on text and image collection analysis - Modeling subjective differences in interpretation - Modeling the changing needs of consumers
In the end, I wish to raise discussion on how these kinds of developments might influence the relationship between art, science, technology and society.
PROF. TIMO HONKELA is currently
a chief scientist at Adaptive Informatics Research Center of Helsinki
University of Technology (TKK). The unit is a center of excellence appointed
by the Academy of Finland. In the center, Honkela is the head of one of
the five research groups called Computational Cognitive Systems. Earlier
he has served as a professor at the laboratory of computer and information
science at TKK and as a professor at the Media Lab of University of Art
and Design Helsinki. He has close to one hundred scientific publications
and has given hundreds of presentations and invited talks in locations
such as Kobe, Beijing, Dubai, Washington D.C., Boston and many European
major cities. Honkela has conducted research on several areas related
to knowledge engineering, cognitive modeling and natural language processing
including a central role in the development of the Websom method for visual
information retrieval and text mining based on the Kohonen self-organizing
map algorithm. Honkela collaborated with George Legrady to produce Pockets
Full of Memories, an interactive museum installation, the concept of which
was created by Prof. Legrady. Honkela is a former long-term chairman of
the Finnish Artificial Intelligence Society. He is currently the chair
of the IFIP working group on knowledge representation and reasoning (WG
12.1) in which position he organized AKRR'05, the international and interdisciplinary
conference on adaptive knowledge representation and reasoning. He has
been in a responsible position for a large number of scientific conferences
and workshops. Honkela's spare time interests include music, go and golf.
Host: Professor George Legrady,
Media Arts & Technology
|
![]() |
![]() |