Re: Lexicon Making

On 4/24/06, veritosproject@gmail.com <veritosproject@...> wrote:
> When you make vocabulary, do you get a text and make what words you
> need or make the words first and then translate?
With my current project I started with the list of semantic primes
from Natural Semantic Metalanguage -- not lexicalizing every one
of them, but making sure I could express each with, if not a root word,
then a no more than two word phrase.  Also I created a handful
of additonal content words like "xejom" (cat).  Then wrote a number of
sample sentences to try out the grammar with, and then took
Rick Harrison's ULD, ran it through a Perl script that added a
column with 99 (a guestimate of inverse frequency), one with codes
for semantic category, and one with "n" (for "no, I don't want a
root word for this") and worked through it assigning lower numbers
to the words I expected I would use, setting their "want root"
field to "y", and corrected their semantic category
field.  I also combined some ULD entries into one, and added
some new entries (for grammatical particles I want that the ULD
takes no cognizance of).  Then sorted by expected frequency, and
used another script to merge these meanings into unused word
shapes in my lexicon (matching up the semantic categories;
this engelang codes the semantic category (substance, process,
quality/quantity, and relation) in the final consonant of the root)).
The lexicon now has about 300 entries (many of them phrases
rather than single words); I'll do some more writing and translation,
coining a few single words as needed, and do the corpus frequency
analysis before I relex the existing vocabulary and add much new
vocabulary.

I think with gzb I started with major vocabulary creation, coming up
with nearly 200 root words in the first couple of months (I didn't
do any automatic vocabulary generation with gzb), while doing
some minor translation and sample sentence writing but nothing
major until a few months later.

--
Jim Henry
http://www.pobox.com/~jimhenry/conlang.htm