On Mon, Nov 10, 2008 at 1:37 PM, Matt Arnold
<matt.mattarn@gmail.com> wrote:
A couple of years ago I recommended to Robin a type of Lojban random
sentence generator that relies on templates. Taking an existing
utterance that parses, you transform each cmavo into the upper-case
symbol for its selma'o. You transform each gismu, lujvo and fu'ivla
into a tag meaning < gismu >, < lujvo > and < fu'ivla > respectively.
You have a template.
To use the template to make a random sentence, replace each selma'o
tag with a cmavo from that selma'o, and each gismu with an actual
gismu. You get a sentence that parses. Let's say you have the template
LE MI GISMU CU GISMU KO'A2. Fill in the slots randomly with "le do
bakni cu sovna di'u" and you get something nonsensical (Your cow is
an egg of the last utterance.), but it parses.
The reason I bring this up now is that I would like to find out which
templates are the most common in the searchable corpus of Lojban
utterances, such as IRC logs. This would suggest very useful templates
for the home-game I'm building with dice, paper, and ceramics. I am
considering recording audio learning courses in which I would group
Lojban utterances (sensable ones) by template, to provide a variety of
examples of simple valid sentence structures, and relate selma'o
through substitution. Perhaps someone could intuit the most common and
useful templates, if not search it with textfile processing.
Smallish point: Is there any reason not to merge < gismu >, < lujvo >, and < fu'ivla > into < brivla >? The distinctions between those three seem not relevant to the stated purpose. { le do .arxokuna cu barkla di'u } works just as well, right?
mu'o mi'e komfo,amonan