Re: Corpora: US company claims patent

Bill Imlah (
Wed, 15 Oct 1997 12:09:45 +0100

John Hutchins wrote:

> To be legitimate a US patent would have to demonstrate the
> 'invention' of a procedure not previously used or described.

It would be sufficient to demonstrate that this was an
improvement of a known procedure. I believe that this is
what the patent is arguing, since there doesn't appear to be a
detailed description of a procedure in the patent itself
(if you don't fully disclose a way to make your invention
work, or don't disclose enough to show how to make it work
given what is publicly known in the field, then your patent is
most likely invalid).

In this case, the essence would appear to be that using a
simple, regularised language as an interlingua is advantageous
over using an irregular, idiosyncratic natural language.

Since people were already constructing such languages for
translation systems in the eighties (ie. semantic representation
languages) there is a strong argument for a lack of inventive step
(ie. obviousness) in this patent: it's the fact that it's a
constructed, simplified language that appears to
be relevant to the invention, not whether people use it to
communicate with each other.

In other words, the use of constructed languages as interlingua
was already known, as were their advantages over human languages,
of simplification and regularisation. Given that, the inventor
would have to be able to argue that there was something about
the use of languages like Esperanto that had some technical advantage
over constructed languages that were not used as international
languages, and that this advantage was not obvious. The words
"rock" and "hard place" spring to mind....


-- Bill Imlah
Canon Research Centre Europe, 20 Alan Turing Rd, Guildford,
Surrey, GU2 5YF, UK.
voice: +44-1483-448844 fax: +44-1483-448845