The purpose of this research is to develop a new voice-communication medium called voiscape. By this medium, two or more persons can talk each other by using a virtual sound space based on spatial sound technology. People can move freely in the "sound room" and people can choose, enter, and exit from a room. Cocktail-party like situation can be created with this medium. In short, this project targets realizing Second-Life-like world by sound.

Complex computer and communication systems require complex policies. I studied methods of combining (composing) policies, or I studied component-based policies. I have also studied methods of (policy-based) QoS guarantee (especially, Diffserv) on IP networks.

The axis-specified search is a method of text search, especially for encyclopedia search. Users (searchers) specify an axis of search in addition to keywords. The search results are arranged along the axis. For example, if the axis is "year", the result is a chronological table. The axis can also be "geography" or various types of "quantity".

This is a small-scale research on developing a method for reproducing and displaying exactly or almost the same Web page as the original (i.e., complete or incomplete reproduction).

What is CCM?

CCM (Chemical Casting Model, or Chemical-Computation Model) is a model for emergent computation. CCM works only with local information, i.e., each reaction (unit action) in CCM refers only a few data, and there is no global plan (i..e., no program) for the computation. CCM is based on a production system (a type of rule-based computation method), which is similar to chemical reaction systems (rather than conventional production systems used for developing expert systems). Two features of CCM are as follows.

• CCM works with locally-defined evaluation functions, which is called local order degrees.
• The order of reactions (rule applications) is random (or, stochastic or non-deterministic). Randomness is a source of emergence.

In addition, catalysts do not change by reactions but affects the reaction velocity, a type of tunneling effect is caused by reaction rule composition, and annealing-like effects are caused by Frustration Accumulation Method (FAM).

What is RACA?

RACA is a model for complex systems. Conventional cellular automata, which was studied by Wolfram, are synchronous and deterministic. On the contrary, RACA are asynchronous and non-deterministic, or more precisely, they are randomized. RACA and CCM are similar in their stochastic behavior. The research of RACA was initially started when I computed cellular automata by using CCM.

The aim of this research was to establish methods of vector processing of logic programming languages and/or symbolic vector processing. The main target machines were vector supercomputers, such as Cray X-MP or Hitachi S-820. However, this research also aimed symbolic supercomputing on SIMD parallel processors, such as CM-1.

I tried to apply methods of linguistics, in which human languages are treated, to analyze programs. I was involved in this research only a short period, i.e., while I was writing my master's thesis. I tried to continue this research, but I got no fruits.

I deveped a programming language called Dihybrid, which was for MIMD computers, as the graduation research of the university. However, the processor could not be completed. After I was employed, I developed several compilation techniques for a Fortran compiler for the SIMD-type supercomputer S-810 and others in a group of several people. The major techniques are methods of dataflow analysis to decide vectorizability, and a method of vector register assignment.

I developed compilers for general-purpose computers since I was in the third grade of the university. The first one is a compiler of SIMPL, which appeared monthly in a Japanese magazine called "bit", but I developed it independently from the magazine article. When I was in the graduate school, I developed a Pascal compiler to be used in the Educational Computer Center of the University of Tokyo. After I was employed, I was involved in two Fortran compiler development projects. I described the first one, a development for vector processors, in another entry. In the redevelopment of compiler after so-called IBM case, we developed a dataflow analysis methods for arrays.