On 16 Aug 2007, at 17:05, Nicolas Roard wrote:
> it's not really a full featured natural language processing, we can
> use an easy to parse grammar...
> The newton was a bit closer to natural language processing, but it
> only did so and worked ok because the set of keywords used to infer a
> request was quite small.
> Usually when you work with a small vocabulary it's not very hard to
> infer proper queries (and discard the rest of the vocabulary that
> doesn't blend in / isn't useful). It works surprisingly well (as in,
> wow the computer understand me, while in fact it's merely a trick only
> possible because of the limited vocabulary).
A 'natural language' parser for a vocabulary the size used in text
adventures is really easy to write (even in FORTRAN). I think it
would be nice if we defined some simple imperative sentence
structures (e.g. '{verb} to {noun} with {noun}') and allowed
components to register nouns and verbs (maybe even noun and verb
phrases) which would then be turned into scripting actions.
The nice side effect of this is that it makes localisation much
easier; you define corresponding imperative structures in different
languages and then someone just needs to translate the noun and verb
list for each application. I'm not sure it would work well for
languages like Japanese, where the grammar isn't tied to sentence
structure, but for European languages it should work well.
David
_______________________________________________
Etoile-dev mailing list
[email protected]
https://mail.gna.org/listinfo/etoile-dev
