The real-time natural language translation telephone is just one of the wonders that lies in store for us courtesy of Japanese technology, according to Professor Hideo Aiso of the Department of Electrical Engineering at Keio University, in Japan. Speaking at the annual computer science conference of the US Association for Computing Machinery in Louisville, Kentucky, […]
The real-time natural language translation telephone is just one of the wonders that lies in store for us courtesy of Japanese technology, according to Professor Hideo Aiso of the Department of Electrical Engineering at Keio University, in Japan. Speaking at the annual computer science conference of the US Association for Computing Machinery in Louisville, Kentucky, Professor Aiso explained that the Automatic Telephone Interpretation Research Project being conducted by the Advanced Telecommunications Research Institute in Japan is one of many computer-related projects being funded by private industry and the Japanese government, reports Microbytes Daily. The telephone project includes work on speech recognition, machine translation in real time, and speech synthesis. Ideal for monolingual people making telesales calls to other countries, the system could come into use as soon as 1992 in a limited way. The aim of the project is to demonstrate such a device with a limited vocabulary of between 1,000 and 3,000 commonly-used business words by then. Other major projects being funded by MITI, the Japanese Ministry for International Trade and Industry include $8m for research into three-dimensional integrated circuits made up of four layers that include organic materials, as part of the national effort to create a neural network system similar to the nervous system of lower forms of life. There is $23m for a project to develop memory chips with access times of less than 10nS – there are already statics that have access times as low as 2nS, but they are expensive and have only thousands of memory cells per chip. The chips are primarily for use in supercomputers. The third stage of the Institute for Computers Of Tomorrow effort involves the creation of a knowledge-based, multi-processor computer capable of surpassing the performance of the current generation of supercomputers; work is proceeding on a 64-processor machine programmed in a hybrid language based on Prolog, but ultimately the machine could have as many as 10,000 processing elements. $100m has been budgeted for this project, with completion set for 1991. There is $20m being devoted to the development of advanced robots by 1990, among them a nuclear-powered robot, a robot designed to operate on the sea-bed, and a disaster-prevention robot. The robots are designed to work where man cannot go; in nuclear power plants, or in dangerous high-temperature disasters such as factory fires involving chemicals, Aiso noted. The robot projects are also contributing to other research, such as in pattern recognition. France and Spain are working with the Japanese in the robotic research. The $25m Sigma project to create standard workstations and a methodology for efficient software development is continuing, and another $20m has been budgeted over seven years to create an interoperable database system and to work towards standardisation of distributed databases. There is $6m for the development of a machine translation system for automatic conversion of the leading Asian languages – including Korean, Chinese, Indonesian, Japanese and Thai, into English. Human Frontier Science The Future Personal Informational Environment Development ’21 project has a $13m budget to create a new information processing system with advanced Japanese-language capability by 1993. And the latest newest national project to be approved, just a few months ago, is the Human Frontier Science Programme, with $80m budgeted over the next 20 years: the aim is to study brain functions with the goal of creating a neurocomputer by looking at biological functions at the molecular level – and the directors have decided to open an office here in London for the research. The three elements of the study will be computational neuroscience, technology, and current neural computer science, Aiso said. The Japanese government is also watching with interest the privately-funded Tron real time operating nucleus development programme, and the Laboratory for International Fuzzy Engineering Research, backed by 46 companies, whi
ch seeks to create the hardware, software, and archi.cw 8 tecture for computers that can handle imprecise information and make intelligent best-guesses. The major application, Aiso suggested, would be in robots that would be able to make decisions for themselves. Among the 46 private investors in the program is the US National Aeronautics & Space Administration’s Johnson Space Center, Aiso noted.