Transmedia machine and its keyword search over image texts

This paper describes a new type of workstation for personal information management of printed materials such as books and periodicals. This workstation is named Transmedia Machine, and has been studied at Hokkaido University since 1983. It inputs and stores printed texts as images to construct a database of printed materials, and provides its users with database search functions. It has a screen text-editor to rewrite input texts as well as retrieved text information. It does not convert input image texts into strings of character codes, and hence requires no character recognition. This means texts are input as they exist without any errors nor any font changes. A new technology was developped to search image texts for arbitrarily given keywords. It does not need to recognize each character completely, and hence the search can be performed in a reasonable time. A prototype machine was implemented in the end of 1985 by software changes of a desk-top-publishing workstation Canon EZPS. The prototype inputs printed texts of books or periodicals through its image reader, and provides users with a screen editor to rewrite input image texts.