University of Tsukuba
Research Center for Knowledge Communities


Formation and Representation of Community Knowledge

In JAPANESE

T O P MEMBER RESEARCH CONTACT LINKS






Research Theme

Identification and Representation of Kodoin and Itoin (ancient metal made stamps)



We are building a database of Kodoin and Itoin (ancient metal made stamps) including their 3-dimensional forms, as a research topic for the virtual museum contents project.




Similarity and Expression of Three-Dimensional Structures


The definition of the similarity among three-dimensional structures is a fundamental element to represent the status of sets of three-dimensional structures. We have been studying on methods to extract functional site of proteins based on graph representation, as a basic research for the protein classification based on the sturcture of functional sites. Pairs of same amino acid residue present in two proteins are repred with nodes, and two nodes are linked when the distance relationship between the amino acids represented with that node are similar. Once a graph structure is constructed, the largest subgraph whose nodes are completely connected (denoted maximum clique) express the group of amino acids residues whose three-dimensional allocation is equivalent. Based on the assumption that amino acid residues with equivalent three-dimensional allocation is required for the expression of similar protein function, amino acid residues extracted from the described method are probably the functional site of the protein under consideration.

We are currently focusing mainly on fast algorigthms to extract maximum cliques, because a disadvantage of this method is a high computational cost, and an efficient algorithm is necessary. An application of the method is the representation of three-dimensional structure of organic compounds, where the amino acids are substituted with atoms, and nodes represent pairs of atoms between two organic compounds. Similarity measure among three-dimensional structure was defined after the definition lf similarity necessary for the extraction of maximum cliques. Results from actual organic compounds indicate that the method is useful as an index of similarity among three-dimensional structure of organic compounds.



Structure of Concepts of Technical and Scientific Knowledge


This research is on the analysis of specialized and technical knowledge from the aspect of concepts represented as technical terms and semantic relationships among therms. Elucidating properties of such structure of concepts is valuable for the representation of knowledge. Current main target fields are Biology and Medicine.

We have built a structure of concepts from the field of molecular biology, using hierarchical relationships. Following aspects of the structure were studied: temporal variation of the structure, alteration of the specialization level of the whole knowledge, mechanism of incorporation of new concepts, and the estimation of the specialization level of content. The results indicate that newly introduced concepts are often specific, which becomes general concepts with time. The tendency of concepts to move to a more high-ranking hierarchy was also clarified. Furthermore, nearly half of concepts that belong to the same hierarchical level are permutated with new concepts. A forumula to estimate the specialization level of content was also developed, which is useful to provide information adequate to the users' specialization level.


Representation and Properties of Chain Structure of Scenes


This research treats the representation of structure involving time. Specific target is stories (fiction, tales, etc.). Stories are defined as a chain sequence of scenes, which are the units of events, and currently four methods of extraction of events are proposed. These methods use different scale of events, providing different viewpoints to represent stories. Sequence between events are classified according to functional and logical types.

Analysis of published stories using all four extraction methods indicated that similar parts of story have similar functional sequence before and after the indicated parts. Both events and functional sequences are useful for the extraction of characteristic story parts.


Ultra High Speed Network Simulation


The network means the model where an object is represented by the interaction among elements. Many physical or abstract entities can be modeled with such a network. High speed simulation of such model is useful to exploit the behavior of the whole network and relationship with the behavior of elements. The analysis is particularly useful for large scale networks, and provides understandings on the phonomena of the whole network, prediction, control and design of networks. We have developed a high speed computer system for network simulation. The simulator executes directly by hardware circuit without the software layer, differing form conventional simulation systems that are software running on ordinary computers. The execution speed is ten thousand to one million times faster than software simulators. This simulator is currently used for the analysis of gene regulation network of biological cells, which is important for the elucidation of biological phenomena and for drug design.


Medical Informed Consent Support System


This research treats how to provide specialized knowledge to non specialists. The clarity or easiness of understanding of an explanation depends not only on the description form, but also on the knowledge level of the person that receives the explanation. Current main target field is Medicine. The system consists of knowledge infrastructure, medical explanation supply unit, biomolecular information supply unit, biological reaction simulation unit, front end computer unit, and user interface unit.



(C) 2007 Formation and Representation of Community Knowledge. All Rights Reserved.