I worked on something similar to this some time ago. But I quickly
realized that we have a language that is very, very good at expressing
this sort of relationship: lojban.
With that in mind, refer you to mlismu, a random bridi generator.
Specifically, it parses a data file written in lojban that contains
relationships of the sort you're describing, and produces bridi that
match the right types.
Your "must-be" and "can-act-as" are actually flip sides of the same
coin. So instead of "x1 must-be agent", I'd say something like
.i ro bajra ka'e gasnu
(Or equivalently, the set of runners is a subset of the set of agents).
Take a look at it, particularly the fatci.txt. It's not thorough at
all, and there are some things I put in there that are known to be
false, but produced humorous results (since the whole thing started as
a joke). Hopefully, I've marked those with {je'u nai} or {zo'o}, but
some may have escaped.
http://students.cec.wustl.edu/~adam/lojban/mlismu/
On Tue, Nov 2, 2010 at 1:14 PM, Oren <get.oren@gmail.com> wrote:
> Re: "Again, the important thing is which individual places accept what sorts
> of arguments. The gismu itself just relates those places."
>
> So then, the concept of my spreadsheet *DOES* contain useful and valid
> information, but would only be complete if it were expanded to include all
> the 'oblique' sumti places as well?
>
> Re:"Does it bother you that *{mi pinxe lo jubme} would also be considered
> semantic nonsense, because tables aren't the sort of thing that one can
> drink?"
>
> If things like "agent/object" are specified in these definitions, why
> shouldn't all 'sensical' general classes like material states
> ("liquid/solid") be included as well? This is in part why I was referring to
> these 'classifiers' as 'tags' originally. As long as people can easily point
> to a construction and say "according to this sumti's implied class and that
> selbri's meaning, this makes no sense," I think that type of judgement
> should have a clear litmus test. And there's nothing stopping us. With a
> vocabulary of less than 1500 words, many of which fall into regular
> sub-classes in the thesaurus, I see no reason why we shouldn't have this
> resource.
>
> So, to expand the scope here, I'm proposing that each and every sumti
> position in gismu definitions list explicit tags for baseline sensicality.
> That is, for bajra:
>
> bajra: x1 runs on surface x2 using limbs x3 with gait x4
>
> Now account for baseline sensicality:
>
> x1 must-be agent...
> x2 must-be material...
> x3 must-be material, must-be movable-part...
> x4 must-be manner...
>
> Now let's envision that these clearly specified 'baseline sensicality tags'
> for sumti positions are like 'keyhole definitions' that only these explicit
> classes can fit. Now, each sumti position also gets any number of 'key
> definitions' for what it can fit into, or what sumti places it can
> sensically 'fill.'
>
> x1 can-act-as agent, can-act-as moving-thing, can-act-as athlete...
> x2 can-act-as general-place, can-act-as surface...
> x3 can-act-as body-part...
> x4 can-act-as idea...
>
> Now, if we do this for every gismu, I imagine we'd end up with many
> high-frequency tags like "agent" and "material," and several hundred less
> frequent tags like "liquid" "body-part." Each of these tags would have a
> list of sumti positions it requires, and a (probably much larger) list of
> sumti positions that can "sensically" fit that semantic role.
> This data/document would not only provide a richer (many-to-many) series of
> 'categories' for vocabulary study lists, there are a series of new
> applications this would allow. You could automatically gauge the degree of
> 'figurative language' used in a text. You could automatically generate
> sensical example sentences for given vocabulary (or even generate a minimal
> spanning sensible sentence for a set of words). You could even develop a
> kind of auto-complete function for a lojban-specific text-editor: as you
> begin to type a sumti in, a list of 'sensical' suggestions could come up in
> a tooltip window. If we get this data, I'd totally code that!
>
> But I want to make sure I'm understanding the nature of this data set.
> Please let me know if I'm still making sense, and if I do, I'll come up with
> technical specs for a web interface to make this data easy to gather and
> manage. Maybe I'll use this as a way to learn to use github.
>> On Tue, Nov 2, 2010 at 12:14, John E Clifford <kali9putra@yahoo.com> wrote:
> co'o mi'e korbi
>
>>
>> Well, xorxes has ideas about how Lojban *does* work, and, with minor
>> exceptions,
>> he has got it right. So Lojban is his "other" language. Sorry you think
>> this
>> discussion is bull-crap; it is trying to work out the ramifications of
>> Lojvan
>> being a logical language, dealing with both the logical part and the
>> language
>> part, and shooting for reasonable resolution where they appear to
>> conflict.
>>>> Sent: Tue, November 2, 2010 9:55:04 AM
>>
>>
>>
>> ----- Original Message ----
>> From: Lindar <lindarthebard@yahoo.com>
>> To: lojban <lojban@googlegroups.com>
>> Subject: [lojban] Re: mi kakne lo bajra
>>
>> Oren, I answered your question some two or three times.
>>
>>
>> Where x2 of broda asks for {nu} and x1 of brode asks for {nu}, {.i
>> broda lo brode} is kosher, because lo brode already -is- an event. For
>> all other cases, an abstractor is necessary.
>>
>> (barring all the other bullcrap/arguments going on right now)
>>
>> xorxes, since you have all these ideas about how Lojban should work,
>> why don't you just make your -own- language and let it stand up to
>> Lojban?
>>
>> --
>> You received this message because you are subscribed to the Google Groups>> --
>> "lojban" group.
>> To post to this group, send email to lojban@googlegroups.com.
>> To unsubscribe from this group, send email to
>> lojban+unsubscribe@googlegroups.com.
>> For more options, visit this group at
>> http://groups.google.com/group/lojban?hl=en.
>>
>>
>>
>>
>> You received this message because you are subscribed to the Google Groups> --
>> "lojban" group.
>> To post to this group, send email to lojban@googlegroups.com.
>> To unsubscribe from this group, send email to
>> lojban+unsubscribe@googlegroups.com.
>> For more options, visit this group at
>> http://groups.google.com/group/lojban?hl=en.
>>
>
>
>
> Oren Robinson
> (315) 569-2888
> 102 Morrison Ave
> Somerville, MA 02144
>
> --
> You received this message because you are subscribed to the Google Groups
> "lojban" group.
> To post to this group, send email to lojban@googlegroups.com.
> To unsubscribe from this group, send email to
> lojban+unsubscribe@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/lojban?hl=en.
>
--
You received this message because you are subscribed to the Google Groups "lojban" group.
To post to this group, send email to lojban@googlegroups.com.
To unsubscribe from this group, send email to lojban+unsubscribe@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/lojban?hl=en.