Re: [lojban] Specifying sumti types: another revision of gimste is complete

2014-05-31 18:48 GMT+04:00 Gleki Arxokuna <gleki.is.my.name@gmail.com>:

2014-05-31 18:27 GMT+04:00 Jorge Llambías <jjllambias@gmail.com>:

On Sat, May 31, 2014 at 4:02 AM, Gleki Arxokuna <gleki.is.my.name@gmail.com> wrote:

I added my semantic categorization to the sheet. It was created long long ago and wasn't meant to be used with tesumti interactions table. So in case there are inconsistencies you are free to correct it.

Also filters are now shown in column headers so you can quickly show only rows with necessary values in a chosen column.

That's useful. You have 7 gismu that don't start the x2 column with "x2", 8 that don't start the x3 column with "x3", 2 for x4 and 1 for x5.

I will check them when others leave the sheet so that I can make corrections without disturbing anyone.

Corrections made.

I still see issues on gismu with properties. For example, "jibni" has x2 (same type as x1), but "darno" has x2 (object). I agree that the x2 has to be of the same type as x1 in these but then objects and events must share the same type (spatio-temporal entities), since events can be close to objects in position.

Two options:
1. Split the definition into:
"x1 (event,object) is near or close to x2 (event, object) in x3 (property); x1 (number, abstraction) is near or close to x2 (same type as x1) in x3 (property)"

This will quickly make definitions bulky. On the other hand there are not so many space gismu.

2. Make objects and events one type which will lead to {lo mlatu ka'e fasnu}.

Any other ideas?

I don't have an issue with "lo mlatu ka'e fasnu".

If so why not consider objects as assertions or even properties? This will quickly destroy the whole system of variable types leaving only their interactions relevant.

We don't usually think of living things as "happening", especially since we have the more specific "jmive" for them, But in any case, I'm now not happy with "same type as" for comparatives. I realized that anything can be compared with anything else if you just choose the right property for comparison. For instance:

li ci zmadu lo sfofa lo ka mi'o xo kau casnu ce'u
"The number three exceeds sofas in how many times we discuss them".

Corrections made.

Yes, there are already similar places in other gismu that dont interact with "same type as..." way.

Since there are properties that can easily accomodate any type (such as "lo ka ce'u se casnu") then x1 and x2 are not really restricted to one of the types of the number/object-event/proposition/property/relation/... typology. It's really the property x3 that fixes the type of x1 and x2 so I'd say: "x1 is near or close to x2 in x3 (property of x1 and x2)".

Notice that "(property of x1 and x2}" is not the same as "{relation between x1 and x2}", which is required by something like "ckini". Properties have a single open argument and relations have (at least) two,

You mean that e.g. bruna3 has one ce'u that is equally applied to bruna1 and bruna2 whereas ckini3 and simxu2 have two ce'u inside?

Note taken.

Corrections made.

So we've got a new type/subtype (whetever you call it) "relation". Such place contains two {ce'u} inside:

porsi2, lanzu3, ckini3, jilra3, simxu2, rimni4.

You can easily find them by searching for {ce'u} on the current sheet.

Is the x2 of porsi an assertion, or should it be a (transitive, antisymmetric, total) binary relation? I think "porsi" works something like "li pa ce'o li re ce'o li ci cu porsi lo ka ce'u mleca ce'u"

It doesn't contradict that porsi2 is an assertion since {ka} is a subclass of {du'u} but I added your example to the sheet (very last columns) until a better glossing of this place is found.

If "property" is a subclass of "assertion" (I'd prefer "proposition")

"assertion" was chosen because it has higher frequency in English. Don't forget that this is to be meant "Simple English gimste" (it is still part of "teach simple lojban" project) although we can have any number of columns for developers.

then they don't belong in the same level of the typology. Shouldn't the types be mutually exclusive?

"assertion" is du'u. property is {ka}. Isn't {ka} the same as {du'u} but with at least one ce'u inside? [See http://mw.lojban.org/index.php?title=ka,_du%27u,_si%27o,_ce%27u,_zo%27e]

For mupli you have "x2 (property of x2)". mupli/fadni/traji/cnano/rirci should have the same for x3, and it should be the same type as x1 (a group of that type). If you have x3 as a set, then property x2 is not a property of x3, but a property of the members of x3.

Fixed. I moved all of them to "General =>Set structure" class, however, steci should probably also belong to them.
as well as some other gismu from "Groups & parts" class.

"selte'i" does belong to the same class. Unfortunately "steci" has x1 and x2 reversed.

For future considerations: "General comparisons" class might also intersect with them and thus needs rearranging.

I think place order is relevant. (thing-with-property, thing-with-property, property), e.g. "zmadu", (thing-with-property, property, thing-with-property), e.g. "mupli". and (property, thing-with-property, thing-with-property), e.g. "steci" (the only one?) are three different classes, although clearly they can all be grouped in one superclass.

Let's not mix "klesi" and "structure" columns then. "klesi" is about semantics, "structure" is about interaction of places and variable type declaration.

Further development and evolution will give those columns more precise definitions.

mu'o mi'e xorxes

--
You received this message because you are subscribed to the Google Groups "lojban" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lojban+unsubscribe@googlegroups.com.
To post to this group, send email to lojban@googlegroups.com.
Visit this group at http://groups.google.com/group/lojban.
For more options, visit https://groups.google.com/d/optout.