Print

Print


Henrik Theiling, On 04/09/2006 23:31:
> And Rosta writes:
>> ...
>> You describe the standard problems with taxonomic vocabularies, but
>> I am of the minority opinion that these problems can be well
>> circumvented if approached the right way. Suppose you have 60
>> syllables. Other things being equal, this gives you a taxonomic tree
>> where each node supports 60 branches. Then you need to slot your
>> concepts into this taxonomy, following the principle that a concept
>> can be assigned to a form of n syllables only when all forms of n-1
>> syllables have been assigned concepts. That solves the
>> deep-and-sparse problem. But it does mean that you can't work from
>> standard language-independent quasi-scientific taxonomies.
>> ...
> 
> The problem I still see here is when to decide that your n-level tree
> is 'full' and you need another level.  In principle, the 'standard'
> approach to create taxonomic vocab does not produce any empty paths
> *in principle*, but only in practice.  I.e., you could go down any
> path in your taxonomic tree and get some meaning, only it is probably
> quite unlikely that all longer words are really used so most paths of
> length n are probably not to be translated when a dictionary for, say,
> English would be written.
> 
> I think in short I want to say: I don't yet understand the difference
> of your technique.

The depth-and-sparseness would arise, I take it, mainly from extending one section of the taxonomy and not others, so tht, say, the taxonomy for actions is broad and shallow, while that for animals is deep. 

You, Henrik, seem to be describing something different: a taxonomic vocabulary concocted according to my recipe, but in which there is no correlation between frequency (of usage) and word-length/taxonomic specificity. This is remediable by settling on a taxonomy that reflects frequency of usage in the first place.

--And.