There is one other very significant factor which I have not mentioned
and this is "truth". If I put "Stefan Boltzmann" or "Black Body" into
Google (or Wolfram) I get out the correct law. If I put 200 deg C into
Wolfram it says (correctly) that the absolute temperature is 473.16K.
You see ice melts at 273.16.
This is not a trivial point. When we translate we look at truth
(consistent of course with a translation). Google is interested, or
claims to be interested, in AGI. Now to be a necessary condition for
AGI is to be able to say something in words (any language)
http://docs.google.com/Doc?docid=0AQIg8QuzTONQZGZxenF2NnNfNzY4ZDRxcnJ0aHI&hl=en_GB
contains the translations. Original Arabic + Google + My translation +
Microsoft. I must damn you with faint praise. Microsoft is even worse,
what is "size ground" for heaven's sake?
Both Google and Microsoft have made all the same mistakes with
formulae and Microsoft a few more.
One word for Ed Bond. Yes this is just about it. It was Arabic -
English though
- Ian Parker
On Aug 17, 6:32 pm, Google wrote:
> Hi Ian,
>
> Our machine translation researchers are constantly experimenting with
> many different ways to improve our translation quality.
>
> You can find out more about what we're doing
> here:http://research.google.com/pubs/NaturalLanguageProcessing.html
I have in fact looked at this as I have Microsoft's research, MindNet
etc.
>
> Thanks for the feedback.
>
> Cheers,
> Josh
>
> On Aug 14, 3:16 am, Ian Parker wrote:
>
>
>
> > Dear Google,
>
> > I am making a very general point about AI and Web 2.0 here.
>
> >http://sites.google.com/site/aitranslationproject/deepknowled
>
> > If you are going to continue running a search engine you will have to
> > generate RDF tags. Now an RDF tag (eg. the Stefan Boltzmann law)
> > should be fed into Google Translate. Part of the disambiguation
> > progress MUST consist of comparing translations with RDF tags. The
> > Stefan Boltzmann law is stated correctly in both Wolfram and in a
> > Google search. How do we disambiguate درجة for example. I say that if
> > preceded by a number it should be translated as "power" with the
> > number going ordinal. Alternatively we can disambiguate with RDF
> > knowing we have a 4th power law.
>
> > - Ian Parker
>
> > On Aug 13, 6:00 pm, Google wrote:
>
