Calculating # of things you can say

Language learning: How to speak Toki Pona, translation problems, advice, memory aids, tools and methods to learn Toki Pona and other languages faster
Lingva lernado: Kiel paroli Tokiponon, tradukproblemoj, konsiloj, memoraj helpiloj, iloj kaj metodoj por pli rapide lerni Tokiponon kaj aliajn lingvojn
Post Reply
janMato
Posts: 1545
Joined: Wed Dec 02, 2009 12:21 pm
Location: Takoma Park, MD
Contact:

Calculating # of things you can say

Post by janMato »

Let's try to count how many things we can say. But first, lets kill the "you can say an infinite number of things because you can do something recursively/repetitively", e.g. I saw the old man. I saw the old, old man. I saw the old, old, old man. I saw a dog and another dog (etc). So excluding repetition, how many sentences are there? And the more interesting question, which sort of word would increase the number of things you can say the most?

S li V e DO prep O (where O is the complement of a prep)

So the number of possible sentence of the above form is roughly

125 * 1 * 125 * 1 * 6 *125

... shoot, that doesn't look right, I'm going to have to look up how to do combinations & permutations.

Another way I suppose is how many relationships can you indicate in a single sentence, among how many participants and how many kinds of participants.

An SVO sentence is 1 relationship (of 125 possible kinds), 2 participants (the subject and object)
SVO + 6 different Preps = 8 participants (each of 125 possible kinds), 7 relationships (one of 125 kinds, and 6 more if we exclude duplicates, like, jan li pali kepeken X, kepeken Y, etc.
S la S, i.e. two maximal sentences... 16 participants 9 relationships

If each participant (subject, direct object, complement of a preposition) was a maximal pi chain, of say N M pi N M, the we have (in a la-chain), 32 participants (max) and 24 relationships.

So, in this sort of model, increasing the content words by 1, say an esoteric word for a chapter of a book, would have no effect on the number of relationships among participants that can be expressed. But subordinate clauses (which would be like injecting a sentence fragment into the place of each possible participant slot) would make the number of relationships you can express explode.

If toki pona was a real language and lets imagine there was some real constraint on morphology and root word count (say brain structure issues in animals), it seems that subordinate clauses would necessarily evolve, maybe as some sort of "ni" strategy that would become indistinguishable from subordinate clauses.

mi lukin e jan ni jan li moli e jan ante. I saw the man who murders.
jan ni jan li moli e jan ante li kama jo e pini pakala. Those who murder others come to a bad end.
janSilipu
Posts: 288
Joined: Tue Aug 09, 2011 9:21 am

Re: Calculating # of things you can say

Post by janSilipu »

Well, your calculations don't take the full complexity of phrases into account -- the number of modifiers and especially of 'pi' modifiers. But more significantly, it does not deals with multi-sentence units, those set up by 'ni:' and those that function like relative clauses already. The limits of these devices has not been tested, although they are probably less agile than English relative clauses, say. But there are probably other tricks to fill many of the gaps..
janMato
Posts: 1545
Joined: Wed Dec 02, 2009 12:21 pm
Location: Takoma Park, MD
Contact:

Re: Calculating # of things you can say

Post by janMato »

The exact numbers aren't so important, just the ranking of the magnitude of change that would be caused by tweaking the various parameters-- i.e. more content words, more function words (prepositions, li, la, e, pi), more slots or making it legal for more things to be stuffed into a given slot (e.g. A en B going into more slots, where AFAIK, it started out as only be allowed for joining subjects)

Adding more content words messes with the "language of few (basic) words" thing and doesn't seem to have the same impact on expressiveness as adjusting the other parameters.

re: complexity of pi
Yeah, I'm not sure how to capture that. moku li lon supa. That seems to describe a small number of situations, semantically-- the food's on the table-- the participants are related to each other, and specifically, touching, probably held together by gravity. But for pi, (and juxtaposition) seems to cover about any relationship that one could imagine.

moku supa li lon. A food characterized by a table (maybe it was placed there? made of it?) exists.
supa moku li lon. A table, characterized by food, exists. (maybe it is made of it, they are somehow touching, the food is a lobster and the lobster owns the table)

pi generates a large number of possible meanings and lon (as a preposition) generates a lot fewer possibilities.

It seems like there should be a way to penalize a language feature that generates lots of semantic ambiguity. Since syntactic ambiguity is easier to count, maybe:

expressiveness = # of things you can say (excluding repeating things) - number of ways something can be parsed
Post Reply