ilo.muni.la: graph how toki pona is used!

[quaint wagon]

https://gregdan3.github.io/ilo-muni/

sina alasa e sona nimi lon ilo Muni la o pana e nasin e sitelen lon tomo ni!

[edgy cradle]

wan

[quaint wagon]

tan seme

[edgy cradle]

heres something i notices

the relative mode is less accurate further back in time

[wicked stratus]

san

[quaint wagon]

https://gregdan3.github.io/ilo-muni/?query=sina+seme+-+sina+seme_3%2C+sina+pali+e+seme+-+sina+pali+e+seme_5&minSentLen=1&scale=rel&start=1470009600&end=1722470400&smoothing=2

[wicked stratus]

I actually have cool graphs I'm just not at home rn 😔

[quaint wagon]

yes, that's true
there's less toki pona going on that far back
the data actually goes to 2010, but barely so, so i ignored it lol

this is why there are multiple scales offered tho

[icy turret]

nanpa wan is roughly twice as common as nanpa tu

[whole epoch]

https://gregdan3.github.io/ilo-muni/?query=nanpa+open%2C+nanpa+pini&minSentLen=1&scale=rel&start=1533081600&end=1722470400&smoothing=1
it's interesting how nanpa open and nanpa pini only started to be used around 2020, with nanpa pini being way more common than nanpa open (probably cause nanpa wan expresses a similar idea)

[icy turret]

indirect evidence for unpopularity of sona-preverb

[edgy cradle]

if 20 people says something and there's 1,000 texts in the month, then that's 2%
but if 20 people say something and there's 100,000 texts in the months, then that's 0.02%

[whole epoch]

indirect evidence for the popularity(?) of preverb open and preverb pini

[novel lintel]

mi anpa lon e kokosila :3 https://gregdan3.github.io/ilo-muni/?query=penpo%2C+penpo+-+ilo+penpo+o+lukin+ala%2C+kokosila&minSentLen=1&scale=rel&start=1659312000&end=1722470400&smoothing=2

[quaint wagon]

there's a related phenomenon where the frequency of the most used words is going down relative to less used words since anywhere from march 2020 to may 2021, continuing to today
(you can see it in that graph of pona above)
i... don't know what this means tho :P

[quaint wagon]

psst, check the absolute mode
there's like 30 occurrences there :P

[icy turret]

damn, its not often that you can attribute a word to one month and then essentially never again

[wicked stratus]

oh gosh it's The Dip™️ mk. 2 /musi

[whole epoch]

i will present my data in the most biased way as to make my opinion seem more correct
-# /musi
but yea fair lol

[quaint wagon]

in general, take anything that has fewer than ~400 occurrences all-time (you can check with cumulative) with a grain of salt
that data can be relevant, but it can be much more easily swayed by errors, or affected by an individual speaker

if you wanna know how powerful exactly one speaker is, look up ilo o ken e toki ni

[whole epoch]

do you think it might be cause the community in general are speaking about more diverse topics? pona, toki and other common words are also more common among beginners/ the start of conversations, so them decreasing in usage might imply that people are having more in-depth cconversations/that "beginner conversations" are getting more uncommon

idk im thinking out loud

[quaint wagon]

essentially yes! i said a similar thing yesterday, altho i described it as more complex topics rather than more diverse topics, but same principle

[whole epoch]

gotcha

that might just be cause the toki pona community is bigger and has gotten less,,, meta for lack of a better term

[quaint wagon]

ehehhe, yep

[whole epoch]

using tp as the medium of conversation instead of as the topic

[quaint wagon]

i am surprised i haven't seen anyone look up ki yet

to be clear, ki's results are nonsense and unavoidably so lmao

there are at least 7 things i'm aware of which ki appears in that are not uses of the toki pona word ki

[whole epoch]

indirect evidence of the unpopularity of preverb open/pini 😔

[quaint wagon]

eeeheheh

[novel lintel]

nimi ki seme pi toki pona :p · ona tu li lon

[icy turret]

@quaint wagon feature suggestion: create a list of commentary matching particular search requests
use it to comment on searches that could be misleading

e.g. if someone looks up san remind them the search is partially "spoiled" by your presence

the counterargument to adding this is we cant possibly account for every misleading search so it might be better done in conversation than in gui

the countercounterargument is that we can account for the most potentially frequent of those

[arctic yoke]

a lot of people, i have noticed, seem to think my name is a toki pona question word /musi

[drifting breach]

i guess most of the speakers (at least the beginner ones) don't even know about this preverb🦦

[round mural]

it's almost like this word died along with the meme

[tall rune]

what was itomi

[tall rune]

ah

[mint knoll]

https://gregdan3.github.io/ilo-muni/?query=linluwi%2C+majuna%2C+jami&minSentLen=6&scale=abs&start=1470009600&end=1722470400&smoothing=2 compared the 3 finalists of utala pi nimi sin

[mint knoll]

https://gregdan3.github.io/ilo-muni/?query=ojuta%2C+mijun&minSentLen=1&scale=abs&start=1596240000&end=1722470400&smoothing=2 ojuta vs mijun

[icy turret]

alasa is growing in relative terms, both as a "standalone" predicate and as a preverb

[fresh sentinel]

how does that compare to lukin preverb?

[drifting breach]

isn't there a special syntax for marking only preverbs or only verbs or only smth?

[icy turret]

there isn't

lukin is harder to estimate for this exact reason

[meager steeple]

#1187212477155528804

la nimi ona li awen lon tenpo mute

[quaint wagon]

^ tenpo kama la mi weka e tomo ni tan sona pi ilo Muni

[quaint wagon]

woah how's it transparent

[brittle mulch]

right click on the graph

and press copy image

[quaint wagon]

WUH????

never knew that was a thing

[brittle mulch]

SINA LI IJO E NI

[icy turret]

yep thats a thing

[quaint wagon]

SITELEN LI TAN ILO ChartJS

LI TAN ALA MI

[brittle mulch]

a

[quaint wagon]

MI PANA E NANPA E NASIN TASO E SITELEN ALA

[brittle mulch]

peli pani

[quaint wagon]

seme

are you saying
very funny
perhaps

[tall rune]

you can see when i joined the community :3

also it strips double letters :c

[quaint wagon]

gonna be real, this will probably never not be a feature (necessity) of ilo muni
there's just Way too many ways to write words if i don't collapse duplicate letters
which makes the database too massive to deliver as i'm currently doing it

[tall rune]

yeah thats fair

[quaint wagon]

there's a similar but less massive problem with capitals

[tall rune]

mhm

[quaint wagon]

fwiw, in processing it's more like

• collapse duplicates (preserving the first cap)
• score
• lowercase
• frequency count
  that is to say, i score things with their caps, then remove caps later for space/counting reasons

[icy turret]

you could solve that by first reading an english dictionary, then refusing to deduplicate letters if they match an english word
but thats frankly silly and unhelpful

[quaint wagon]

1. other languages
2. phonomatches in english
3. phonomatches in other languages

these are really all the same problem

this also assumes the english words are, themselves, rendered appropriately

[icy turret]

okay. replace "solve" with "improve"

[quaint wagon]

i'm not honestly sure it would even do that

not without a pretty sizeable manual processing step, anyway

[tall rune]

apparently there was a little stella bump in june 2021

[icy turret]

my hypothesis is that 1 is essentially not affected, and 2 3 are fewer than the number of english words that are correctly preserved by this

but like yeah

its so far removed from the point of the tool that its just not worth your time

[meager steeple]

"mi li" and "sina li" can still be found in sentences like "soweli mi li pona"

[icy turret]

unfortunately this can't be used to draw conclusions about the frequency of ungrammatical li because what ilo tani said yeah

[meager steeple]

sona mi la mun li wile ken e ni: ^ mi li la ni li alasa e toki lon open toki taso

[icy turret]

@quaint wagon check this out

two bits of insight from this

1. 90% of toki pona is those words (i think is how youre meant to read it???)

[quaint wagon]

nope the scale is wrong

[icy turret]

fuc

[quaint wagon]

the shape is right but the numbers are wrong

my bad

[icy turret]

right, its the same scale problem

[quaint wagon]

check again in relative; you'll have exactly one line anyway

[icy turret]

anyway insight two is not affected by that issue

[quaint wagon]

20% of toki pona is those words, btw
roughly

[icy turret]

2. assuming early tokiponists weren't using really fancy toki pona which they weren't, the fact that these words add up to less probably means that early toki pona data has more non toki pona noise in it

[quaint wagon]

oh yeah no doubt

to help a bit, move up to sentlen of 2

or even 3

[icy turret]

minor gui nuisance: you might not want to allow users to do this

[quaint wagon]

the math will be done against the total number of words, not sents of len 3+, but more words = more opportunities to score correctly

[icy turret]

ye

@quaint wagon vertical lines grid in the graph background. are they hardcoded

[quaint wagon]

nah they're a default of chart js that I did not investigate

i could do a ton more presentation wise

[icy turret]

noted

@quaint wagon whats the correct way to compare the frequency of a vs a a vs a a a

i tried a - a a, a a - a a a, a a a - a a a a but from the vibe of the chart i feel like ive not considered something

misali has been declining in mentions

[fresh sentinel]

oh, the

[icy turret]

apeja li mi peaks during every sptp because we're a crowd and we shout it

pretty expected. had to use log scale to see toki Inli taso

[drifting breach]

the unpa season

[merry walrus]

I am going to listen and see what you are trying to say

Oh I've listened to þis before

[merry walrus]

I listened to þe whole song really trying to interpret þe lyrics in a sexual way and it feels so much like a stretch þat I don't þink þis is what your talking about or maybe I am just really misunderstandign what she is saying

Oh þere are lyrics

I'll come back

[icy turret]

in case youre in doubt, read what acronym the album name "utala ni (li) pona a" spells out :p

[quaint wagon]

the best you can do is limit the sentence length and deal with that limited perspective
subtracting the number of a two up is actually more accurate, but more limited
otherwise, i still errantly double count in this context

this is a mistake google makes too, if their docs are to be trusted

[icy turret]

sona pona

[merry walrus]

Þis is really ugh

I really misunderstood her

Like misheard her words a lot

I loved þis song because it sounded like

two lovers who are apart woefully missing each oþer

But now þat I'm reading þe lyrics

It's just like

not þat

ugh

Anyways I want to find out what Majeka was being referred to here

Oh I found it

No?

Þere are instances of majeka as a magic nimisin in 2021

Like 3 times which ig aren't here

or maybe I'm dumb because I don't know how þis þing works

[drifting breach]

do these spikes really exists?

[merry walrus]

mi ni e mama sina li is powe

[novel lintel]

nimi "AAA" li lon toki [🇨🇦] :p

[short narwhal]

a tomo ni li lon

[novel lintel]

nanpa li lili mute a · o lukin lon nasin [Absolute]

[short narwhal]

mi kalama e sitelen akesi
(there are slight rounding errors due to how i made the wave in audacity: 1, normalizing to be within +/-1; 2, rounding to 1 decimal point; 3, each month was 1 sample point of a wave, so i slowed it down in audacity to auto smooth it)

[brittle mulch]

all spiked oct 2021 hmm

[merry walrus]

It seems þat (ik pi li is an old grammatical þing) people's grammar's getting better

[uneven prairie]

but this nanpa tu includes nanpa tu wan and nanpa tu tu

[fresh sentinel]

i'm also curious about

• pi * e
• li pi

[icy turret]

data volume too low to be reliable

but for li pi we can at least say it was probably more used pre pandemic

and pre pandemic you legitimately had a lot of people learn from jan Pije so

then again jan Pije's site already purged weird usages after pu came out

[fresh sentinel]

pandemic probably meant that lots of people were learning for the first time with good resources

[icy turret]

not necessarily

jan lentan's came out in like what 2021? i forget

someone should check tbh

[fresh sentinel]

good = avoiding li pi

what about pi * e? like "soweli pi moku e kala"

(i can't run Muni because it doesn't work in my browser at this time, i've filed a GitHub issue)

[icy turret]

the thing is that ilo Muni doesn't do a straightup regex-like search over the text, it finds appropriate trigrams

and there are few trigrams that fit pi * e

most are too infrequent to be included

so the result is unrepresentative

and you cant sum them up, it shows as different lines

[fresh sentinel]

ahhh

[icy turret]

so for example

if there were 2 results for pi walo e

in all texts

ilo Muni doesn't store that trigram because not enough data

and won't display it here at all

it also only displays the top 10 ish(?) when doing a wildcard search

[icy turret]

@quaint wagon smoothing suggestion: try a kernel with soft edges

google ngrams doesn't do that but it feels like itd be better at literal smoothing of the graph

so instead of peaks becoming plateaus, they would become more spread out peaks

-# ofc this comes with the disclaimer that this feature request is for whenever you feel like working on muni again

[quaint wagon]

I replied btw, with a few sites you can test to investigate what component is the broken one

[pseudo smelt]

https://gregdan3.github.io/ilo-muni/?query=mu%2C+kalama&minSentLen=1&scale=rel&start=1470009600&end=1722470400&smoothing=2
mu li kama suli tawa kalama

[mint knoll]

https://gregdan3.github.io/ilo-muni/?query=pu%2C+ku%2C+su&minSentLen=1&scale=abs&start=1470009600&end=1722470400&smoothing=2

pu ku su

[quaint wagon]

I wonder if I could deliver the given graph as the open graph image for the site, if the URL params are filled in

That's... Probably irresponsible with my database? I actually don't know what it looks like networking wise when you fetch the metadata of a link
Is discord doing that for you and then sending you the result? Or is it each individual who sees the link?

[tall rune]

it does it for everyone individually

theres this one website whose embed is a simple math problem, except its randomly generated every time, so when you post it, everyone sees something different

[tall rune]

its still an interesting graph tho

[short narwhal]

pi * e = pie 🥧 👍

[fresh sentinel]

ooooh

[quaint wagon]

Ooo that's super neat
And actually brings up a different question:
I probably need to have an actual server to do this trick, huh?
All my JS is client side lmao, so that idea is not happening
Although it does occur to me that an alternate way for me to deliver the app would be to have the entire thing be hosted on like, Vercel? And keep the entire DB in memory, using just sql.js on the server side
Something to investigate for later

[tall rune]

https://txnor.com/mathchallenge

this is the thing

[quaint wagon]

LMAO

[mint knoll]

7

[tall rune]

nah its 19 trust

[quaint wagon]

actually, this is a search i could allow
the reason I limit it to standing in for a single word is because if you search for the top 10 matches of a given form like this, you'll only get back 10 phrases of the minimum matchable length for the search; shorter phrases are also more numerous.

granted there are some places around funky grammatical features that could be exceptions, but there won't be many of them

[fresh sentinel]

ooh

[quaint wagon]

what i could do is performance testing on allowing multiple wildcards; there are few enough terms available that it could work

[whole epoch]

it's interesting how, despite interjection lon becoming more unpopular, interjection ni has stayed at a steady level

also cool to see how "mi mute" for "we" is on a steady decline
back when i learnt tp originally way back when, it was almost universal to use it (as i remember it anyways)

[icy turret]

lon-yes was quiiite commonplace

its been backlashed

[whole epoch]

yeye

i just expected ni have gotten more popular

[quaint wagon]

@icy turret i think it was you who posted it but i can't find it; you posted the comparison between nimisin and a few of the other "standard low use but around" type words like linluwi and majuna? and i have thoughts about that as well, actually
well, one thought really:
a user of the word nimisin is much more likely to talk about lower use words; conversely, those who don't use the word nimisin talk about newer words less, resulting in the phrase nimi sin being about as used as the word nimisin.
in fact, this is something i could probably demonstrate in my primary db? by counting the number of distinct authors who have sentences containing nimisin, versus the number of distinct authors who have sentences containing the phrase nimi sin

[fresh sentinel]

i am generally curious about like
if Muni describes a word as popular
is it because many people are using the word, or because spiders Georg is using the word a lot

[quaint wagon]

i know for a fact this is happening to lipamanka and nano, because their names are primarily said by themselves
but determining that currently requires checking manually

you can be extremely confident that words over rank 150 are actually in use by a variety of speakers, but the range of confidence drops significantly after that since you go from several thousand uses at rank 150 to several hundred uses at rank 200; that's an amount which you could swing by in a single day of being silly

[fresh sentinel]

idk the privacy implications of this but
if Muni could say whether a word is being used by under 10, 100, 1000, 10000 people
and then lines could have different thickness or opacity depending on the category
that would maybe make it easy to visualize if a spike is a silly spike

this wouldn't catch mu mu mu

[quaint wagon]

ooooo wait that actually is interesting and i don't think it would be hard to check

uagh it would be a lot of queries tho lmao

well, not on ilo muni's side
db generation side

i'm unsure how to represent it on the graph tho; line thickness gets difficult to judge if there are more than 3 distinct thicknesses

[fresh sentinel]

3 might be enough? for 10, 100, 1000

[quaint wagon]

also, for reference, distinct author count is imprecise bc of pluralkit and more generally bc i can't combine authors across platforms

[fresh sentinel]

idk how often a word will be used by more than 1000 people

[short narwhal]

jan mute ala li kepeken nimi Nimisin. ni li tan ike nanpa tan ni: nimisin Georg li kepeken nimi ni mute la nanpa li ante ike la ona o lon ala nanpa

[fresh sentinel]

in one month

[quaint wagon]

oh i see, you want that info on a monthly basis

that could be harder

that makes sense tho

[fresh sentinel]

not necessarily! this is just the first idea that came to me

and wouldn't change the UI too much anu seme

[tall rune]

we found a spiders moment yesterday

[icy turret]

@quaint wagon you could create some kind of measure for how concentrated word use to one person vs many people

[icy turret]

this sounds vaguely similar to the gini coefficient

[quaint wagon]

my god

i have no way to fix something like this

like, not reasonably

[tall rune]

o tonsi tawa ale

[quaint wagon]

idk at least it's clear it's a spike of silliness

[icy turret]

o tonsi tawa ale

[whole epoch]

same thing for lonsi but a group of spiders

[short narwhal]

sina alasa e pipi Georg

[fresh sentinel]

this would show up as an under-10-people line in my proposal, which may help

[quaint wagon]

right

[tall rune]

makes sense

[icy turret]

you know, i can ruin your data by posting procedurally generated text at least once a month in large quantities

[quaint wagon]

i would omit you, personally, from the data in that case

[icy turret]

the correct solution is ofc to ban me from ilo Muni

ye

[quaint wagon]

actually i think i neglected to mention this entirely but i do omit #jaki from the data

[icy turret]

yeye

[fresh sentinel]

reasonable

[quaint wagon]

which i think is completely understandable ye

uagh i've been fiddling with postgres all morning at work and now i'm looking at the sqlite db

explodes due to slightly different keywords

[fresh sentinel]

starting a public server and calling the general chat #jaki /utala

[tall rune]

other stuff i did yesterday was making these sorts of graphs to compare the relative popularity of two words, pretty fun to look at

[quaint wagon]

you can do that! ma pona's jaki is the only bot channel i omit bc it's the largest channel on the server and it isn't even close

[icy turret]

worse you can create a thread and invite ten people to spam it with plausibly real text

[quaint wagon]

that channel, in text, is 4gb of the 56gb of the entire raw discord dataset

[fresh sentinel]

holy shit lmao

[quaint wagon]

i'll grant that text is a lot of spare discord fluff i.e. authors and their roles and metadata on a per message basis
but.

[tall rune]

i also learned that nobody said mije in november 2016

[quaint wagon]

why do you have a-a on the graph?

[tall rune]

on the previous ones it was to highlight the 0 line

[tall rune]

i just didnt bother removing it for this one

[icy turret]

no one on reddit did cause reddit was small and discord communities didnt exists yet

[tall rune]

makes sense

[quaint wagon]

yeah, until i think march 2017? the data is all telegram and reddit

[fresh sentinel]

a time when men didn't exist

[quaint wagon]

mije is reliably less used than meli which is very amusing to me

[short narwhal]

tenpo pi mije ala

[tall rune]

this graph shows meli vs mije and yeah, meli is almost always more popular except for the weird spikiness at the beginning

[icy turret]

im guessing this is cause demographics

[tall rune]

which is probably just because of the small volume of data

there was a time in early 2023 tho where mije was more popular

[icy turret]

i just saw someone say minisin and this is making gears turn in my head

[short narwhal]

misinin

[quaint wagon]

gears grinding violently, making a terrible screeching noise, but just barely turning

[icy turret]

what would a minisin look like

[short narwhal]

me when i do this again

mi ni sin

[quaint wagon]

ok i think i can do author counts on a per ngram basis

[icy turret]

look into the gini coeff genuimely

it might be better

[quaint wagon]

would i not need author count in order to get that info

the sqlite db side of it would be easier than i thought initially, since i can just add another row to the frequency table "num_authors"; the number of authors of a given word or phrase is always containable in the same dimensions as a frequency entry i.e. some phrase, some min sent len, and some date (representing a range)

in order to count authorship, i need a way to know what distinct authors have said a given term, for which i can use their edgedb generated uuid (note: authors are joined by their combo of platform and platform id+name, and only may be entered when i enter a new message) and probably stuff that in a per-entry set for the lifetime of the ngram counter, which is the time it takes me to count a month of data
when i query sentences, i just tack on the additional info of sentence.message.author.id and use that to update the authorship sets corresponding to the phrase i'm updating
and at the end of a month, write back the length of each set side by side with the frequency data

[icy turret]

i think it might just be num authors of all ngrams per month? if you count authors for every ngram individually you lose the 0 case

not sure tho, maybe thats stupid

[quaint wagon]

i am intent on getting num authors of all ngrams per month and num authors for each ngram

the frequency table and ranks table are nearly identical for a reason

that reason is, while you can count the ranks info from the frequency table, it takes over 20x the reads

i'm essentially packing a more specific type of query into that table

[quaint wagon]

question:
if i implement a selector for smoothing method, would that be overwhelming
there's already. a lot of options lmao

@icy turret @fresh sentinel @whole epoch @tall rune

[fresh sentinel]

i haven't actually used the tool yet ehehe

[quaint wagon]

ah fair

did you see my suggested search btw
it really shouldn't take more than 3s to resolve a simple text search even with the default settings
and since those other tools all work, it must be something I'm doing that's breaking on your browser
idk what tho, and you can't pull up dev tools on mobile, aaaaaaa

[icy turret]

i was the one to propose it so i have to say no
I would call the options soft kernel (smooth) and hard kernel (google ngrams style) which should indicate to people why theyd ever want to use the hard kernel (they don't), but maybe you have better naming conventions in mind

[quaint wagon]

uh, my naming method would have been to use the name of the smoothing method as i discover it on Wikipedia pages about stats w.r.t . timeseries data

[whole epoch]

wouldnt mind it so go for it if you wanna do that
it like, wouldnt get in the way of anything except a tiny bit of screen realestate on the top

and if it's too much you can just not use it (assuming the default is good)

[quaint wagon]

@wanton shoal (hi) suggested exponential smoothing and it's pretty good bc it mostly preserves changes in direction while moving the peaks and troughs nearer to one another; it also passes the tonsi and misikeke tests, correctly showing 0 for any point prior to the initial non-zero point of those graphs
i have also found gaussian and median, which are respectively extra curvy and extra blocky; gaussian fails the tonsi/misikeke tests but is good for trend analysis; median does decently at those tests, passing them up to 30 smoothing

actually median technically fails the tonsi test bc of the 3 occurrences in July 2019, 0 in Aug/sep, and 21 in Oct; it omits the July data for a while

[wanton shoal]

might be good to add a dropdown where u can choose from multiple methods? and maybe do a little explainer on what they do and don’t do, for the uninitiated?

[quaint wagon]

ye, a drop-down was what i was suggesting here

[whole epoch]

that's basically what's done for the different charting modes and i assume it'd be the same for smoothing

in the help page there are explanations

[wanton shoal]

oh am on phone and didn’t scroll up that far soz xd

[quaint wagon]

np!

[quaint wagon]

granted I think omitting tiny early data is a better failure mode than extending early data to earlier than the word existed

[wanton shoal]

i played around yesterday with only including data up to time T, but not beyond it, so data wouldn’t get extended into the past

however it didn’t seem to work all that well tbh

[quaint wagon]

exponential?

or median?

[novel lintel]

nimis in pi toki [Mini] anu seme

[wanton shoal]

normal sliding avg

the thing is, it works to some extent, but imo the data only barely resembles the input haha
here's the input:

and 5 smoothing applied

i can see the resemblance, but in the original the peaks aren't nearly the same height (i suppose the second peak is flattened a lot)

(also, "the tonsi test" is a very funny phrase to me for some reason)

[wanton shoal]

median smoothing also sucks

destroys the data, yet suffers from the same issues as simple avg

mi sona ala. pilin mi li ni: ken tu li lon.

1. ilo pona ni li awen. ona li pona ala, taso ona li sama ilo Ngram
2. ilo one-sided exponential li kama kepeken.

... 3. some data analyst helps out :D

lots of research into things that deal well with cyclic data, passing only certain bands for signal processing, whatever
but i feel like just stat vis isn't really a focus of any research. very thin info online

taso mi toki taso. jan Kekan San o toki e pali :D

[quaint wagon]

mi wile pana e nasin ken mute li wile pana e ilo Exponential lon open tan ni: linja sin li sama linja open

[wanton shoal]

lon

[quaint wagon]

ken suli la nasin ona li pona nanpa wan tawa lukin pi tenpo pini

taso, mi awen alasa e sona li wile toki tawa jan pi sona nanpa, a a

[wanton shoal]

lon. mi pilin e ni: ilo wan li pona lon ijo p# wan. ilo tu li pona ala lon ona li pona lon ijo p# tu.

[quaint wagon]

aaaaa ken

[wanton shoal]

lukin la ilo pi ken ale li lon ala :D

sama ijo ale hahaha

[quaint wagon]

a a a, lon

[merry walrus]

Out here doing þe Lord's work 🙏

Ignore þe Belarusians

hijksafkjajg

[quaint wagon]

.... belarusians?

[merry walrus]

Þe only oþer example of pokasi I've seen is it being used for Belarus

[pseudo smelt]

… pokasi?

[quaint wagon]

anyway this increases the necessity of unique author tracking ehehhehe

[merry walrus]

Yes

How much have I contributed to þe Pokasi population

Probably most because I made it up

Þe greatest þing I ever did was make a nimisin in my first SP sentence

I didn't even know what a nimisin was

[wanton shoal]

power move

[pseudo smelt]

wawa

[wanton shoal]

https://github.com/gregdan3/ilo-muni/pull/15 hehe

[quaint wagon]

i am halfway through writing this as we speak LMAO

[wanton shoal]

also re: issue #13 regex having look-behind and look-ahead is cursed anyways, makes it a CFG

lmao

[quaint wagon]

fair enough tho, this is easy to slot into my work!

[wanton shoal]

i was faster >:)

[quaint wagon]

ehehe

[wanton shoal]

would have been a good 20 mins faster still if i didn't accidentally work on the unforked branch lmao

[quaint wagon]

lmao

i should probably do my work on branches instead of directly on main now huh

[wanton shoal]

👏

preach

even when working alone, i've noticed that as soon as a project grows in size and i wanna do some quick fixes, PRs are still a godsent

cuz manually selecting what to include in the diff is annoying, when you've refactored half a file :D

[quaint wagon]

true, ehe

lazygit makes it smooth when necessary tho

i very rarely have to dip into my git plugin's diff view to fix things

[wanton shoal]

i guess it's also a result of me refusing to use the command line for git anymore, since i only work in IDEs nowadays, and it's just so much easier for me xd

so i'm kinda accustomed to select all, commit, push

but that does hide the fact i forgor to checkout another branch haha

[quaint wagon]

ehehe, np

i've merged your branch to a separate branch merge-smoother which i wish i just called smoother; gonna add a smoother url param and some logic to disable the button as necessary!

aside: oh god i have no idea if my js is any good, i am not good at computer

ah jk i just missed that you did so in reading the github diff!

[wanton shoal]

u made a thing, which is more than a lot of people can say ab themselves! :D

...but yes, some refactoring might be good haha

[quaint wagon]

yeahhhhh

ehehhehe

see the thing is
sona toki is the best code i've ever written
aaaand this whole project was downhill from there :D

i do know at a minimum that the sqlite file is a mess bc i've mixed the responsibilities of querying and mutating the data there
and that the input file sucks bc i'm doing a bunch of crappy splitting instead of like, actually parsing user input

[meager steeple]

btw, are you analyzing each message or sentences within a message?

if i put a full stop or semicolon would that turn the message into two sentences

[quaint wagon]

sentences within a message

my sentence tokenizer isn't perfect
quotes count for it too
altho i am considering removing them honestly

[meager steeple]

sona

[quaint wagon]

i wanna sit down and do some analysis to that end
on one hand,

• quotes, double and single, are definitely used to distinguish sentences
  on the other hand,
• in toki pona, they much more often already are marked as sentences based on the other punctuation
  ofc, both of these are anecdotal statements- i wanna know which is more true and when

also, i wanna do intra-word punctuation in the word tokenizer for isn't don't
not because i need that to tokenize toki pona, but because doing so would slightly increase my accuracy of detecting toki pona

[quaint wagon]

@wanton shoal turns out that exponential is misleading for peak-y data! Fook

[merry walrus]

I don't get what þe smooþing does so I'll assume it's doing good here

[quaint wagon]

the data is really noisy to begin with, so smoothing attempts to reduce that noise while still representing the data faithfully

but the problem is, under some circumstances the smoothing is inaccurate

• if you have data with peaks at specific times, the peaks can be spread out in a misleading way (center averaging turns them into plateaus, and exponential smoothing turns them into slopes)
• if you have data that is zero for a while then starts, smoothing can spread the data to before there was a real data point

ok having the realization that i can't really produce an appropriate smoothing algorithm without knowing what it means for there to be a "signal"

really, the smoothing necessary just depends on the input
and different terms will have different signals

median works extremely well on continuous data, but demolishes seasonal data instantly

extremely funny

no smoothing for comparison

exponential is reliably the least silly of all the smooting methods

relative minmax is nice for comparison multiple test cases at once

[merry walrus]

damn misikeke barely edges it out

RIP tonsi

[quaint wagon]

lmao

[tall rune]

im skeptical of smoothing and just turn it off when i use the tool usually

[hidden tide]

tenpo Mopiju li lon

as you can observe, in times of tenpo pi kama sona la troughs form

here's the usage trend of various video games that were played on ma pona, i ought to add gartic phone but i don't know what the common tokiponization of it is

[mild glacier]

hey

hows this?:

[hidden tide]

each instance of a larger string of letters adds to the count of the smaller strings here

[mild glacier]

ah okay, cool!

hows this?

[hidden tide]

snazzy

[meager steeple]

oo that little jump in lu near the end is because of april fools day when everybody spoke tuki tiki

[pseudo smelt]

a!

muti

[hidden tide]

muti mute

wait no i'm blending languages

muti a

[pseudo smelt]

muti muta
(sona mi la “muta” li nimi majuna pi toki sike)

[meager steeple]

muti tu

anu seme

[hidden tide]

i mean i guess

but i think just a afterwords sounds more natural

anyways, here's jan Wano's graph that looks at each string individually without accumulation

take away the first line and we can see the trend of how many as are used for laughter

[quaint wagon]

A AA A A A A

[merry walrus]

gyuhjkeafafd

Damn þe rest of tp þrough L just barely beat out þe particles

[quaint wagon]

LMAO YOU DID ALL OF THEM?
send link? @merry walrus

[meager steeple]

sina wan a li suli e "ijo sama"

[quaint wagon]

MI WAWA

[merry walrus]

It's only þrough L, I just wanted to see how long until þe particles got outpaced (ignoring a because it would give any side an unfair advantage)

https://gregdan3.github.io/ilo-muni/?query=li+%2B+e+%2B+o+%2B+pi+%2B+la+%2B+en+%2B+kin%2C+akesi+%2B+ala+%2B+alasa+%2B+anpa+%2B+ante+%2B+anu+%2B+awen+%2B+esun+%2B+ijo+%2B+ilo+%2B+ike+%2B+insa+%2B+jan+%2B+jaki+%2B+jelo+%2B+jo+%2B+kala+%2B+kalama+%2B+kasi+%2B+kama+%2B+ken+%2B+kepeken+%2B+kili+%2B+kiwen+%2B+ko+%2B+kon+%2B+kute+%2B+kulupu+%2B+lape+%2B+laso+%2B+lawa+%2B+len+%2B+lete+%2B+linja+%2B+lili+%2B+lipu+%2B+loje+%2B+lon+%2B+luka+%2B+lukin+%2B+lupa&minSentLen=1&scale=cmsum&start=1280620800&end=1722470400

[meager steeple]

wawa...

nimi ale li lon ala a a

[quaint wagon]

i know this for a fact: the pure particles are only about 20% of toki pona

[merry walrus]

I did it from memorie so I might be missing a couple words in L

[quaint wagon]

you haven't included "toki" or "pona" or "sona", or any of the pronouns for that matter which will ofc be a huge portion of all toki pona

i get that was the point to be clear :P

[merry walrus]

Oh I missed kule and laso damn

[quaint wagon]

those won't make much of a difference ehehe

[merry walrus]

Critical words

[quaint wagon]

i mean number wise

[quaint wagon]

i found the source of the wan/tu/mute/luka/ale spike in oct 2021

there is a channel in one server where people genuinely counted to the high thousands in the pu numbering system

well, i say people, but it seems to have been almost entirely one person

[merry walrus]

based

[meager steeple]

aaa musi

[quaint wagon]

things brings on an entirely new question
like, do i exclude that

how much is that reflective of "using the language"

[merry walrus]

Anoþer reason for a unique auþor feature

[quaint wagon]

you know, that's true tbh

i'll leave it be for now

[placid crow]

@quaint wagon how did you get the data for Muni? Ik it was through online to communities, but how did you download the data itself?

[merry walrus]

Illegal meþods /idk

[quaint wagon]

i talk about that on the about page! In short, Reddit has an archival project, Telegram lets you export chats, and Discord you'll have to read on the site

[low timber]

toki pono vs pono

[verbal marten]

[mild glacier]

ah mi sona :D

what are the little numbers that people add?

like "a a a_6

[whole epoch]

that's the minimum length needed

[mild glacier]

ah okay

[whole epoch]

so like
toki_4 only shows toki from phrases with 4 or more words

[mild glacier]

ahhhh... okay!

i get it!

gants :3

[whole epoch]

it's very useful!

[tall rune]

i found another interesting spike

[whole epoch]

lmao??

[tall rune]

(this is a relative graph, the absolute graph is less extreme)

still there tho

[mild glacier]

hows this?

[whole epoch]

too little to say anything tbh, only a handful of uses

[mild glacier]

yeah

lol :3

(i have a weird urge to say sorry after every problem i kinda make)

[tall rune]

for some reason the tool isnt letting me search "unpa li ken *"

i wanna find out what the next word isssss

[whole epoch]

just unpa li ken has no results found so that's probably why

[tall rune]

thats incorrect

maybe youre in a too specific timeframe

"unpa li ken" is the big spike in march 2016

[mild glacier]

musi la:

[whole epoch]

huh, yea now it works for some reason??

nasa a

[tall rune]

the spike is some 6 word phrase that starts with unpa li ken

[mild glacier]

what is it?

[tall rune]

idk

its not letting me do the *

[mild glacier]

unpa li ken pakala e sina e sina kin

:D

[tall rune]

its not pakala

[mild glacier]

ah

[tall rune]

i cant search for it because its not from this server

[whole epoch]

since it's 2016 i assume it must've been like, a meme or running gag on telegram?

or maybe reddit

[tall rune]

i guess

[quaint wagon]

that's a "spike" up to 5 occurrences tbf :P

[tall rune]

ah

it just sticks out funkily

[quaint wagon]

granted it's odd it happened all at once, but probably makes sense to whatever discussion happened there

yeah, there was just a lot less conversation going on at the time

[hidden tide]

[hidden tide]

big spaghetti time:
here's my relative minmax graph attempting to convey the trends in focus over time of various meme phrases throughout ma pona's history. by "meme" phrases, i mean any word or phrase that is often repeated among speakers or has some sort of server addon like an emoji or a sticker. obviously incomplete. o lukin pona a!
https://gregdan3.github.io/ilo-muni/?query=a+-+a_2%2C+omekapo%2C+omekalike%2C+ale+li+pona%2C+mu%2C+kijetesantakalu%2C+ikea%2C+ilo+nanpa%2C+kekan+san%2C+jan+telakoman+li%2C+misali%2C+lonsi%2C+mi+tawa+tomo%2C+mi+tawa+e+tomo%2C+lon+-+lon+a+-+lon+ala%2C+lon+ala%2C+mi+sona+ala+a%2C+pingo%2C+kasi+ike+mute%2C+usawi%2C+nimi+sin%2C+nimisin%2C+tonsi%2C+su%2C+owe%2C+kamalawala%2C+wawa%2C+waken%2C+akesi+kule%2C+lon+a&minSentLen=1&scale=normrel&start=1470009600&end=1722470400&smoothing=0&smoother=cwin

something interesting that can be found by max smoothing it out is how, depending on what shape is portrayed, "wawa" is in a steady trend upward and either has passed over or is on the verge of passing over "ale li pona"

[meager steeple]

for standalone sentences it's been surpassed

[hidden tide]

fun

[fresh sentinel]

https://gregdan3.github.io/ilo-muni/?query=wawa+-+wawa_2%2C+epiku+-+epiku_2&minSentLen=1&scale=rel&start=1470009600&end=1722470400&smoothing=2&smoother=cwin

i feel like wawa interjection is maybe the closest pu equivalent to epiku interjection

that's how i use it anyway

[icy turret]

also pona a

[fresh sentinel]

oh good point

is it possible to search for a two-word interjection?

[icy turret]

pona a - pona a_3

[river shard]

uh
wawa_2 ?

[meager steeple]

wawa_2 is all messages that contain wawa with length 2 or more

[fresh sentinel]

https://gregdan3.github.io/ilo-muni/?query=wawa+-+wawa_2%2C+epiku+-+epiku_2%2C+pona+a+-+pona+a_3&minSentLen=1&scale=rel&start=1470009600&end=1722470400&smoothing=2&smoother=cwin

[meager steeple]

so if i do wawa - wawa_2, it's the number of sentences containing wawa minus the number of sentences containing wawa of length 2 or more; aka how many times wawa was a standalone sentence/message

[hidden tide]

i added a few more memes and changed the wawa parameters to have it be standalone, it drastically drops in terms of relativity (naturally because it is a word that people say outside of commending someone)
https://gregdan3.github.io/ilo-muni/?query=a+-+a_2%2C+omekapo%2C+omekalike%2C+jipi%2C+slape%2C+jans%2C+ale+li+pona%2C+mu%2C+kijetesantakalu%2C+ikea%2C+ku%2C+nja%2C+yutu%2C+osu%2C+ilo+nanpa%2C+epiku%2C+tuki%2C+kekan+san%2C+jan+telakoman+li%2C+misali%2C+lonsi%2C+mi+tawa+tomo%2C+mi+tawa+e+tomo%2C+lon+-+lon_2%2C+lon+ala+-+lon+ala_3%2C+mi+sona+ala+a%2C+pingo%2C+kasi+ike+mute%2C+usawi%2C+nimi+sin%2C+ojuta%2C+a+pilin+ike%2C+o+uta+e+sike+mi%2C+toki+a+-+toki+a_3%2C+moku+wapa%2C+apeja+li+mi%2C+mi+unpa+e+mama+sina%2C+o+luka+e+kasi%2C+ma+masina%2C+mijun%2C+nimisin%2C+o+moku+e+kala+pona%2C+sutopatikuna%2C+o+moku+e+kala+ike%2C+o+wawa%2C+tonsi%2C+su%2C+owe%2C+kamalawala%2C+wawa+-+wawa_2%2C+waken%2C+akesi+kule%2C+lon+a&minSentLen=1&scale=normrel&start=1470009600&end=1722470400&smoothing=0&smoother=cwin

[winter thicket]

jsyk a link in þe help article is missing a slash

[hidden tide]

for whatever reason, August 2022 was the month of mu

absolute chart tells it better

[quaint wagon]

this is because of sptp!

[quaint wagon]

thanks! fixed it

well, fix is publishing in like 1 minute

[tidal knoll]

https://gregdan3.github.io/ilo-muni/?query=kepeken+pi%2C+kepeken+e%2C+kepeken+lon%2C+lon+kepeken&minSentLen=1&scale=rel&start=1470009600&end=1722470400&smoothing=2&smoother=cwin
nasin pi te kepeken e to li kama lili anu seme

[quaint wagon]

ona li lili mute, taso alasa sina li pana ala e sona ni

[tidal knoll]

nasin pi te kepeken pi to en
nasin pi te kepeken lon to en
nasin pi te lon kepeken to li awen nasa

^ tan fucking seme la ilo pi toki pona taso li pipi e mi?!

[quaint wagon]

o alasa e ni: kepeken *, kepeken ala *
ni la sina ken lukin e ni: nasin ni li lon; ona li suli tawa nasin wan mute pi nasin ante. taso ona li ijo wan lon poka pi nasin ante mute li lili wawa tawa ona ale.

[quaint wagon]

• nimi te en nimi to li lon ala tawa ilo pi toki pona taso tan ni: lipu Linku la ona li lili.
• ona li lili lon sitelen la nasin ale ante la ilo li lukin ala e ona
• (o sona e ni: nimi te li lon tawa ilo Muni, taso nimi to li lon ala tan toki Inli)

[tidal knoll]

https://gregdan3.github.io/ilo-muni/?query=kepeken+toki%2C+kepeken+e%2C+kepeken+ilo%2C+kepeken+nimi%2C+kepeken+nasin%2C+kepeken+ala%2C+kepeken+tenpo%2C+kepeken+sitelen%2C+kepeken+ona%2C+kepeken+ni%2C+kepeken+ala+e%2C+kepeken+ala+nimi%2C+kepeken+ala+kepeken%2C+kepeken+ala+ilo%2C+kepeken+ala+ona%2C+kepeken+ala+nasin%2C+kepeken+ala+toki%2C+kepeken+ala+sitelen%2C+kepeken+ala+ni%2C+kepeken+ala+la&minSentLen=1&scale=rel&start=1470009600&end=1722470400&smoothing=2&smoother=cwin kepeken e li suli nanpa tu

[quaint wagon]

lon! ona li suli nanpa tu, taso ona li open e poki. ona li suli nanpa tu lon poka pi nimi ale ni: open poki en ijo poki li lon.

sina wile alasa e ni:
nimi kepeken la ni li ale ala. taso, nasin "kepeken e" la ni li ale.
https://gregdan3.github.io/ilo-muni/?query=kepeken+e+%2B+kepeken+ala+e%2C+kepeken+toki+%2B+kepeken+ilo+%2B+kepeken+nimi+%2B+kepeken+nasin+%2B+kepeken+tenpo+%2B+kepeken+sitelen+%2B+kepeken+ona+%2B+kepeken+ni+%2B+kepeken+ala+nimi+%2B+kepeken+ala+ilo+%2B+kepeken+ala+ona+%2B+kepeken+ala+nasin+%2B+kepeken+ala+toki+%2B+kepeken+ala+sitelen&minSentLen=1&scale=rel&start=1470009600&end=1722470400&smoothing=2&smoother=cwin

ni la sina ken sona e ni:

• nasin pi nimi e li kama lili
• nasin pi nimi e ala li kama suli
• nasin pi nimi e ala li suli mute tawa nasin pi nimi e

[tidal knoll]

wawa

https://gregdan3.github.io/ilo-muni/?query=mi+e%2C+sina+e%2C+ona+e&minSentLen=1&scale=rel&start=1470009600&end=1722470400&smoothing=2&smoother=cwin
mi e, sina e, ona e
transitive pronouns li lon

https://gregdan3.github.io/ilo-muni/?query=li+mi+e%2C+li+sina+e%2C+li+ona+e&minSentLen=1&scale=rel&start=1470009600&end=1722470400&smoothing=2&smoother=cwin though nobody has ever said li sina e

[fresh sentinel]

ilo penpo li pipi tu e sina

[quaint wagon]

seme a?

[fresh sentinel]

pipi utala tu li lon li sama

pipi kala wan li lon li wile

[plucky gazelle]

(cw: flashing lights) mmh rendering soup

[wicked stratus]

what did that one person say about forbidden spaghetti

[river shard]

Hmmm something about HAVING HIGH STANDARDS?

but I try not to be upsetti

[quaint wagon]

i should Probably prevent people from doing this for exactly the reason you're seeing
but I also don't think it matters enough lmao
maybe it does to avoid bandwidth problems

[hidden tide]

so why did mun Kekan San choose to say walo loje rather than loje walo?

[quaint wagon]

ona li kule ante

[low timber]

nimi suli li suli
nimi sewi li sewi lili
taso, nimi suwi li suwi e sewi ona
nimi seli li awen lili sama seli moli

pakala ni li seme a

ni kin li nasa

[tall rune]

theres a lot of these

both in #sitelen-pona and in the now archived #spt-jaki

@quaint wagon would it be worth excluding that channel?

i remember we talked about how identifying webhooks (which is what these are) didnt work out for you

but excluding spt-jaki from the data seems like a good idea

[tall rune]

cumulativeeeee

[quaint wagon]

hmm!
this is a case where i think the following two improvements would be better:

1. scanning through all my webhook messages for pluralkit messages specifically, so i don't count we hooks
2. scoring sentences with consideration for the message they're a part of

i would like to avoid poking more channels out of the list, preferring a code based solution that more correctly separates good and bad sentences, so that i can keep the intended fairness of the system

in this case, a document level score would ideally mark that first sentence low because it's among a bunch of non-tp words

also, there are other weird spikes that would be fixed by document level scoring! like pipi Kewapi saying "ilo o ken e toki ni a" genuinely some thousand times which i think is absolutely hilarious

[tall rune]

if you can do that, great!

i just thought cutting out that channel would be an easy solution

it feels like a channel where not much would be lost anyway

[quaint wagon]

that's probably true, yes
but i need an excuse to do the webhook check and the document scorer anyway :P

[tall rune]

fair

[mild glacier]

sitelen pi ilo Muni /musi /literal

[pseudo smelt]

musi a
seme a la ona li ilo tawa

[mild glacier]

mi sona ala /shrug

¯_(ツ)_/¯

[icy turret]

@quaint wagon sona pona wants the ilo muni logo cc-licensed btw

talk to em for details idk shit

[mild glacier]

couldnt they just ask for permision to use the picuture?

sorry for asking

i dont know alot of things

[pseudo smelt]

mi kama lukin e sitelen ilo… la

[meager steeple]

pakala...

[mild glacier]

among ku!!!

[novel lintel]

ken a · taso jan [Juwan] li wile cc e ale tan,, ijo pi sona mi ala

[fresh sentinel]

ona li wile pana e sitelen ale tawa poki Wikimilija

[hidden tide]

i believe in waso supremacy (not accounting for soweli which is giant in comparison)

[meager steeple]

aa seme la kijetesantakalu li suli sama akesi sama pipi

[quaint wagon]

psst, the relative log is a bit misleading rn because it shows the actual logarithm of the value rather than changing the y axis to be logarithmic
(it's like this because i was running out of time)
you should post the actual relative graph of this

the shape of the relative log graph is ultimately correct, but the numbers don't mean much

[hidden tide]

(wink wink)

[plucky gazelle]

pretty linjas

[royal orchid]

tan tomo https://discord.com/channels/301377942062366741/1187212477155528804 la, ni:

[royal orchid]

over the past year, anu seme has been, on average, 65% of all instances of anu

[quaint wagon]

sonja put ilo muni on tokipona.org! which is super exciting.

[quaint wagon]

nasin ni kin la sina ken lukin e ni!

[royal orchid]

sona a

[mild glacier]

a okie :3

[neat bramble]

32/4+15 = 19 23
(i can haz more?)

[short narwhal]

lmao i got 24/4+13=19

wait fym 32/4 + 15 = 19 thats 23

[neat bramble]

oh i'm silly

that's what they mean when they say 99% fail

hey look it's txnor the home of the unpa hack

[edgy cradle]

woa what's with that random jump in anu toki

[royal orchid]

sama nimi wawa la, nimi sona kin li kama mute:

[royal orchid]

wan la, ni

musi la, ni

[pseudo smelt]

seme a

[quaint wagon]

ilo li toki e ni:
"tenpo wan la... toki ni li lon."

[drifting breach]

what is "XXX_n" ("a_3", "sona_2", etc.)?

what are + and -?

@quaint wagon

[quaint wagon]

the number after the underscore means to only count the times that word occurred in sentences with that many words or more! e.g. toki_2 counts only times where toki was in a sentence with at least 2 words.
and the plus and minus are literal- add or subtract two or more things. the math being done is, for every period of time that is represented, add together all the words. or, add/subtract, from left to right, to be exact.

[merry walrus]

Kokosila isn't ÞAT far ahead

Huh

Stupid þought

but I wonder how much affect I had on þat huge isipin spike in Oct and Dec

Oh it's a bot reposting some nimisin

And maybe Giggity Mantis yapping about using isipin

[fresh sentinel]

bots aren't counted in ilo Muni

[pseudo drift]

lmaoo

is there any info on how the data is handled here, by the way? I couldn't see any info about how or if it was anonymised

[royal orchid]

ni li tan seme a?!

kalama lon sptp???? toki jaki ala la, mi sona ala e tan ken ante

[whole epoch]

didnt notice this until now but you should probably add somewhere that wildcard ignores e and li

[royal orchid]

nnnnn...
nasa mute a

[whole epoch]

erm seme a

[royal orchid]

pakala nasa a

ona li pana e nimi pi mute nanpa 11-19??????? tan seme a??????? 😵‍💫

[quaint wagon]

Uhhhhhh
It should not
What the fuck

[whole epoch]

nasa a,,

ok now it doesnt but it did before

[quaint wagon]

That's a lil cursed and given the query I'm doing is incredibly simple and the processing I'm doing after that is also simple, I think I can only blame the library I'm using to query the DB?

And it's a funny one so fair

[whole epoch]

know why it would skip the top 10?

[quaint wagon]

Uh, not really, no
The query as written should fetch the top 10 items, because it searches the ranks table and orders that by occurrences
The only way that could return something other than the actual top 10 would be if the query were convinced of having the top 10 items early/incorrectly

[low timber]

ni li ken musi Mapo

musi Mapo la ilo li toki e ni

[quaint wagon]

ILO

[quaint wagon]

I KNOW WHY THIS HAPPENED

there's a funny but unavoidable oversight in how i detect bots vs webhooks
i don't actually have specific info about whether a given author is a webhook
all i know is whether they're a bot, and what roles they have
webhooks never have roles, while bots do if they're in the server
this is all i have to distinguish them as far as i'm aware?
so if i encounter a bot user with no roles, i mark them as a webhook

anyway mappo is not in the server!

if i just do the pluralkit message check thing i can fix this but i have not yet

[royal orchid]

a, sona!

[quaint wagon]

note that when a bot is in the server, they're obligated to have at least one role- their own, which they create on joining, and which cannot be assigned to any user beside that bot

[merry walrus]

pog

[novel lintel]

mi alasa e linja sama kepeken ilo Nanpa

mute la nasin [UCSUR] en nanpa suli li lon · taso ni kin · seme a

https://gregdan3.github.io/ilo-muni/?query=o+sitelen+lon+lipu+sina%2C+o+sitelen+lon%2C+lipu+sina&minSentLen=1&scale=rel&start=1470009600&end=1722470400&smoothing=2&smoother=cwin

suno wan la toki tu wan ni li lon · ma ni ala · ma pi kama sona ala · la ma seme

ni li musi wile mi

wile mi la mi ni ala https://xkcd.com/1138/

xkcd: Heatmap

There are also a lot of global versions of this map showing traffic to English-language websites which are indistinguishable from maps of the location of internet users who are native English speakers.

mi kepeken nanpa pi mute kepeken la linja ale li sama mute tan ante tenpo pi suli kulupu · taso mi alasa e linja pi sama mute mute · la ni ala anu seme

[quaint wagon]

ma wan li lon tawa ni: sina sitelen lon lipu sina
ona la ilo li kama lon tenpo suno ale li toki e ni

[quaint wagon]

mi lukin e ni lon tenpo ni: mi kama sin lon tomo!

[quaint wagon]

ilo ni li lon nasin Webhook
ilo Pulaki kin li ni la mi ken ala sona e ante
nasin tu wan li ken:

• toki pi nasin Webhook o lon
• jan pi ilo Pulaki o lon ala
• mi pali e ilo la mi pana e toki ale pi nasin Webhook tawa ilo Pulaki li alasa e sona ni: ni li tan sijelo seme pi ilo Siko?

mi sin e sona lon tenpo kama la nasin nanpa tu wan o lon anu seme...

[fresh sentinel]

sona mi la, ilo Pulaki li weka e sona mama pi toki majuna

mi sona ala e suli tenpo

taso mi alasa e sona pi toki majuna la ilo Pulaki li toki ala

[quaint wagon]

ona li ni ala
tenpo ale la sina ken pana e sitelen ❓ tawa toki pi ilo Pulaki
ni la ona li toki e sijelo open tawa sina

[fresh sentinel]

n, pakala mi li ken tan ijo ante

[novel lintel]

mi ante e ilo li weka e nasin [UCSUR] e nimi lili e toki pi nanpa taso

mi kama lukin e ijo pi tuki tiki e ijo musi mute ala · ni li musi lili

[novel lintel]

---

ken la ilo o moku e toki anpa tan sitelen pi ma [YouTube] · pali li lili tan ilo sewi [yt-dlp] · taso mi sona ala e ni → mute li pona ala pona

[low timber]

sona kiwen

[quaint wagon]

ni, kin, li sona kiwen

[quaint wagon]

ni li ken wile, taso pali li suli ike
sona mi la mi wile ni la mi o kama jo e toki tan sitelen ale
ni la mi o jo e sitelen ale...

[novel lintel]

sina ken jo ala a e sitelen

sina o sona e sitelen wile · taso poki pi kalama musi li lon

mi ni

yt-dlp https://www.youtube.com/playlist?list=PL3meDZ0v1E3e5hwSyfz9Os9ZUsMw4ecwz --get-comments --exec pre_process:"del %(id)s.comments.json" --print-to-file "%(comments)j" "%(id)s.comments.json" --exec "type %(id)s.comments.json" --skip-download

la ale li pona

a taso kalama pi mute ala li lon poki · la n

"ytsearchall:toki pona" li ken

[quaint wagon]

mi toki ala e sitelen suli li toki e sitelen wile
mi o jo e ona ale
alasa mute a

[novel lintel]

sina jo e ale pi ma mute anu seme · la o alasa e nimi linjuwi pi ma [YouTube] lon ona

[quaint wagon]

aaaaaAAAAA

wawa

[icy turret]

i still don't know the sona kiwen joke

[novel lintel]

-a mi toki ike · mi wile toki e toki anpa pi jan ante e toki pilin · e ni ala → jan li sitelen nimi e toki pi sitelen ona

ni nanpa tu li pona kin · taso lili a

[plucky gazelle]

ni kin li olin Juli

[quaint wagon]

there is not even a joke anymore, it's basically a way to irritate lipamanka
at two separate times, I defended the idea that somebody could use the phrase sona kiwen to refer to something which is difficult to understand
once in toki pona, where the response was middling
the second time in English, the true place where nasin discussion lives, and my defense bugged lipamanka so it argued against my point for a few Hours
mind you after like 20 minutes i was barely involved anymore, but as these things go the discussion remained for a while
eventually most the participants were so annoyed that bringing up sona kiwen genuinely irritated them
naturally pipi Kewapi and I think this is hilarious

[quaint wagon]

sina lukin e toki pi kili Juli!

[quaint wagon]

mi sona! mi kin li wile e toki anpa kulupu

[plucky gazelle]

aaaa

#toki-ale message kepeken ni pi nimi Juwi li lon ala tan seme

[quaint wagon]

nimi li wile lon ilo la ona o mute mute
(mute li nanpa)

mi weka e nimi ale pi mute ni ala tan ni:
ona li mute wawa li anpa e ken ilo

[plucky gazelle]

a

[icy turret]

wild

[drifting breach]

what is that "SITEŚ"?

[tall rune]

site's

[plucky gazelle]

x/sites/

[charred patrol]

/siteɕ/

[quaint wagon]

ilo pre_process li wile seme lon ni?

[novel lintel]

mi,, sona wawa ala

mi kama jo tan lipu pana mi

ken la --write-comments --skip-download taso li pona sama

[quaint wagon]

sona, lukin la ni

sitelen ale la ona li kama jo e toki

wawa

lukin la pali ale ante li wile poki e toki taso

[neat bramble]

wawa

[quaint wagon]

musi la, toki ~luka luka tu tu mute ale li lon lipu ni taso:
#pana message
taso, mi kepeken ilo alasa la toki pi mute wawa li lon...

mi o kepeken nasin nanpa ni anu seme:
toki wan li lon kulupu lili la toki luka tu li lon kulupu suli.
mi open e alasa lon tenpo luka pini.

[devout jetty]

poka pi ma [sanpansiko] la ilo tawa li lon • nimi ona li ilo [muni] a

sike Kapo li lukin li ton(Sí?) ↩️

[Reply to:](#1272180068721889290 message) musi a
seme a la ona li ilo tawa

[novel lintel]

"mute ale" · seme a · toki,, tu ale mute mute mute mute ale anu seme · ni anu ni ala la wawa

[pseudo smelt]

a! musi

[quaint wagon]

(luka luka) (tu tu mute) (ale)
sina ken sona e poki tan kama lili nimi

[novel lintel]

mi awen sona ala :p

[quaint wagon]

pakala

[plucky gazelle]

ni li ike nanpa

[quaint wagon]

tu tu mute li mute mute mute mute
luka luka li lili tawa ni la o suli e ona kepeken mute mute mute mute
ni o suli lon tenpo ale. ale li nanpa.

(5+5) * ((2+2)*20) * (100)

[novel lintel]

taso seme li pana e sona ni → tu en luka ala li lon poki mute

[quaint wagon]

nanpa lili li lon poka pini pi nanpa suli

[novel lintel]

luka tu wan mute li seme

[quaint wagon]

a. mi kama sona e pakala.

[wanton shoal]

mi pilin nasa tan ni

taso mute pi toki "anu seme" li 1820 taso la...

ken

[quaint wagon]

ni li nasa seme?

[wanton shoal]

mi pilin ni: toki "kxk" li lon mute

[quaint wagon]

aaa sona

[wanton shoal]

taso toki "ken ala ken" li mute ala kin lon lipu nanpa la mi ken sona

[quaint wagon]

toki mute pi nimi tu wan li mute lili taso

[wanton shoal]

lon, taso mi pana e ona tawa poki la ona li lukin lon nimi ale toki, lon ala lon?
taso.. ona li suli ala

[quaint wagon]

lukin a

kin la toki lili "kxk" li lon lipu nimi pi ilo alasa

[wanton shoal]

lipu seme?

[quaint wagon]

a, ilo alasa li ilo Sona Toki. ona li kama jo e toki li alasa e sona ni: ni li toki ala toki pona?
lipu ni li toki pona nanpa wan tawa ilo
nimi pi nasin kalama pona li nanpa tu
nimi ijo li suli lon sitelen open li nanpa tu wan
nimi pi sitelen ale ken li nanpa tu tu
nimi mute li lon ala ni ale la, toki li pona ala

[merry walrus]

Damn a completely fucking blows it out of þe water (I wonder how much oþer languages have affected it)

[quaint wagon]

Obviously not none, but given the absolute graph for a follows a similar trend to every other word in the top 20 most frequent, I feel extremely confident its frequency is accurate to its actual occurrence in Toki Pona!

[merry walrus]

pog pog

[tidal knoll]

https://gregdan3.github.io/ilo-muni/?query=o+e+soweli+ala%2C+o+ala+e+soweli&minSentLen=1&scale=rel&start=1470009600&end=1722470400&smoothing=2&smoother=cwin

[icy turret]

i prefer ala la o e soweli

[tidal knoll]

ala la o e soweli: No results found for this query.

[icy turret]

sadness

[meager steeple]

ala la o e soweli

mi toki e ni la ona o kama lon ilo pi tenpo kama

[icy turret]

-# ilo pi penpo kama

[meager steeple]

pakala nimi wawa

[novel lintel]

kon san li kama lon lape mi · kon pi penpo weka · kon pi penpo lon · kon pi penpo kama

[fresh sentinel]

khdhkdsgg

[icy turret]

ona li kon kekan san

[icy turret]

....apparently russia(n(s)) died in 2019

also oof @ that peak around early 2022

apeja consistently peaks every august and i love that

[tall rune]

why is that

[thick gust]

sona musi

[mild glacier]

lon

seme li nena ni?

[fresh sentinel]

which one is red and which one is blue? for both of these

[icy turret]

lon is blue, ala is red

[icy turret]

people singing the lyrics of apeja li mi in chat on suno pi toki pona

[tall rune]

a

[fresh sentinel]

and this one?

[icy turret]

this one wasnt actually a comparison, rather the same trend for both words

but anyway the first entry is blue, the second is red

[fresh sentinel]

ahh

[icy turret]

the ✨ hierarchy ✨

[fresh sentinel]

it would really help if you include the labels on these, most of us don't have the color order memorized

[icy turret]

improved the hierarchy

[icy turret]

ordered from top to bottom

[fresh sentinel]

pona

[icy turret]

i expected namako below core words but nope, right there with noka and nena

[fresh sentinel]

noka and nena probably experience the monsi effect

[quaint wagon]

they do, as far as the charts show, yes

[plucky gazelle]

seme li tu e linja meso

[icy turret]

sina toki e ni anu seme

[plucky gazelle]

ni

[icy turret]

jan li kepeken ala ->
log(0) = undefined ->
sitelen li ala

[plucky gazelle]

a

tan li nanpa.

[icy turret]

tan li nanpa.

[merry walrus]

Why is þat profound /gen

Oh it is explained later down

sorry

[icy turret]

no worries

🐸 🎮

[plucky gazelle]

why does lija even say gaming in that song

like what does it mean

[hidden tide]

she's a gamer

[verbal marten]

What's the monsi effect?

[fresh sentinel]

direction words and body part words are used less frequently in text and VC, and more frequently IRL and in VR

so they show up less frequently in ilo Muni

which only looks at text

[edgy cradle]

first i thought "what, that's crazy" and then i thought "oh yeah that makes sense"

[wicked stratus]

mijun

[wicked stratus]

wait thats cool

what happened ,,, what happened in august 2023,,,,

[hidden tide]

me

[quaint wagon]

jan Mijun li kama lon

[verbal marten]

ni li tan seme?

[verbal marten]

a sona

[quaint wagon]

taso, ni li ale ala

mi alasa

[quaint wagon]

mi lukin e ni lon tenpo pini li weka e sona

[merry walrus]

lmao

[short narwhal]

o tonsi tawa ali (:<

[pseudo smelt]

Be all moving nonbinary-ers

[merry walrus]

📠 but wiþ a t before þe x

or

in þe middle

fa>t<

[thick gust]

what's this?

ah wait, this is without smoothing

[quaint wagon]

uhhhhhhhhh, perhaps that copy/paste blurb where lon is nearly every word in the sentence?

will check db and report back when able

-remindme 7h30m o alasa e lon lon

[latent zodiac]

Set a reminder in 7 hours and 30 minutes from now (<t:1725130846:f>)
View reminders with the reminders command

[thick gust]

ooh, mysterious

[icy turret]

try looking up lon lon lon lon etc in ma pona

im guesing its someokne playing around with "bubble wrap spoiler shields"

[mild glacier]

lipu ni li pu?

[plucky gazelle]

#jaki message
jan li musi sitelen ilo!

[mild glacier]

aa mi sona

[quaint wagon]

aaaaaa
taso, ni li lon tomo jaki ala anu seme?
ilo Muni li lukin ala e tomo ni

[plucky gazelle]

lon ala

[quaint wagon]

pakalaaaa

[rancid quarry]

i cant find the link a

[quaint wagon]

https://gregdan3.github.io/ilo-muni/

[rancid quarry]

thanks

[verbal marten]

ni anu seme?

[quaint wagon]

niiiiii

[pseudo smelt]

nasa

[latent zodiac]

Reminder for @quaint wagon

Reminder from YAGPDB

o alasa e lon lon

[pseudo smelt]

alasa li pini (anu seme)

[quaint wagon]

a, pini

[verbal marten]

taso kulupu ni pi nimi "lon" li tan seme a‽

[quaint wagon]

lukin la ona li musi sitelen kepeken ilo

[verbal marten]

a sona nasa

[pseudo smelt]

a

[thick gust]

turns out they're playing some sort of chess

sona musi a

[merry walrus]

lmao

[mild glacier]

i have an idea now

imma make chess :D

[verbal marten]

toki. mi Sam. mi li pana wile en sona ijo toki pona.

[mild glacier]

sina sona ala sona?

[icy turret]

have we ever talked about the tenpo ni la dropoff

[quaint wagon]

a lot of words and phrases exhibit a significant change in usage starting in 2020, and while it's pretty easy to see the correlation with the pandemic, there's no telling what exactly that implies about the language
i would say something like, a sudden increase is skillful conversation? tenpo ni is a pretty simple and even first day of learning sort of construction used for basic conversation
try contrasting with "tenpo ni la sina pali"?

[icy turret]

it goes too low usage

but yeah

this is a similar dropoff but slower imo

oh fuck nvm, wrong smoothing

similar yeah

[icy turret]

a a mi kala.

[wicked stratus]

a a mi kala

[wicked stratus]

hmm i wonder whats missing

[placid crow]

poki wesi - reddit containment center

[verbal marten]

ma siko pi toki siko

[low timber]

so it’s lipu Wesi, but ilo Siko

[quaint wagon]

@icy turret

10 smoothing, because it's much more apt here

[icy turret]

oh shit, new sources? youtube i assume?

[quaint wagon]

YouTube, and the old forum + its archive of the yahoo group
the data is very sparse there, but it will be available soon!

[icy turret]

awesome

[quaint wagon]

once again asking for any archived irc conversations lmao

they certainly don't exist beyond the few well known examples, sadly

[icy turret]

you know how tokipona.net had a corpus
was it just jan Kipo's corpus or

[quaint wagon]

i don't know tbh
i think it was though

I saw a 2017 fb post where kipo referred to that corpus with a lot more detailed information than anyone but its author would have

[icy turret]

whats responsible for higher traffic in 2007 and 2010 btw

[quaint wagon]

no clue!

this is another instance where i spent a long time running tests and manual queries to be sure i wasn't making some error

for 2010, perhaps the newness of the forum itself?

it came out in Oct 2009

[low timber]

a

[meager steeple]

seme la ona li jan ala

[quaint wagon]

tenpo weka la ona li soko

[meager steeple]

aaaaa sona

[low timber]

kapesi..

nasa a

kapilu

nimi kapesi en nimi kapilu li suli lon tenpo sama

nimi kap- li wawa lon sike 2021

n, nimi kapa ala

[plush gust]

definitely been done before, but i find it very interesting that people are saying "o kama pona" more now vs. just "kama pona"

jan mute pi tenpo pini li toki e "kama pona" tan seme? mi la "o kama pona" makes more sense. maybe it's a difference between announcing somebody has "come well" vs. telling somebody to "well come" but msa

[low timber]

my uneducated guess is that people are trying to steer clear of calques, and kama pona looks like an english calque for "well come"
perhaps "o kama pona" is more tokiponalike. though i don't see a problem with "kama pona" being an exclamation- though nanpa tu la it could be considered a lexicalization

i don't know

[plush gust]

hmmmmm, interesting

[plush gust]

to complete the set

[thick gust]

interesting

[low timber]

something happened mid 2020

[plush gust]

the global pandemic, along with other factors, caused toki pona to have a surge in popularity around then

[low timber]

but this is relative data, right?

[plush gust]

lon

[low timber]

so that shouldn't affect the words relative to each other

maybe the absolute data

[plush gust]

taso toki pona li ante mute lon tenpo ni

[low timber]

ken

[plush gust]

with the influx of people, nasins ante'd mute

[low timber]

i arrived some time after that, so i don't know how different the community was before then

[plush gust]

and specifically phrases like "kama pona" were popular because new people were coming lots

[low timber]

hmm but i'm not sure why that would have caused people to prefer one or the other

[plush gust]

mi kin li sona ala

[icy turret]

overall "monthly volume of toki pona" seen by ilo Muni:

there are high plateaus in 2007 and 2010 and we don't know what they correspond to, as of now

so 2017 represents the first time toki pona escapes this average level of activity, by virtue of reddit + discord

unless, ofc, the communities not yet in muni (like obvs all the irc chat logs which may or may not exist) reshape this graph starkly

@quaint wagon any reason it goes back to march 2002 instead of august 2001 btw?

in the sparse data pre-2017, mi + sina vs li show some amount of inverse relation, which to me sounds alternating between mostly chatting (hence 1st 2nd person) and prose (hence 3rd person)

whereas by now the community is large enough that the movement cancels out

[plucky gazelle]

🔵 mi + sina 🔴 li

[icy turret]

..right fair

do we have any known posts from even earlier?

regardless of ilomuniability

[quaint wagon]

https://web.archive.org/web/20220808015056/https://wyub.github.io/tokiponaarchive/2001-08-08.html
this, archive of an archive,?

[quaint wagon]

facebook could

idk its exact active years but it may make a difference in the 2014-2019 range

[river shard]

facebook groups is a pain
the API didn't work back when facebook had it normal for groups, and now they restricted it further

[icy turret]

i look forward to mKS singlehandedly creating a working facebook scraper

[quaint wagon]

it would be the second

unfortunately, the first is proprietary

altho granted my work wasn't singlehanded on that first one but it was A Lot

[icy turret]

interesting

btw honrstly

like

is the reason for not using the proprietary one legal or academic (repeatability)

cause the contemporary fb group probably generates a negligible ampunt of traffic and having just a dunp that never gets updated is sufficient for muni

[quaint wagon]

Both of these

I can't use it for private purposes, and nobody else would have access to it

[icy turret]

alright fair

[fresh sentinel]

it's repeatable because someone else can just write another facebook scraper /musi

[icy turret]

anu-predicate

[quaint wagon]

nimi o li lon poka la seme?

[icy turret]

a pakala mi

[wide pawn]

ken la o lukin e "anu ala e" kin?

[icy turret]

ive done that on a separate occasion but it wouldn't count towards "anu e" so

[hidden tide]

here are all the pu words placed into ilo muni! have fun.
(don't click if your ilo can't handle the huge amount of data)
https://gregdan3.github.io/ilo-muni/?query=a%2C+akesi%2C+ala%2C+alasa%2C+ale%2C+anpa%2C+ante%2C+anu%2C+awen%2C+e%2C+en%2C+esun%2C+ijo%2C+ike%2C+ilo%2C+insa%2C+jaki%2C+jan%2C+jelo%2C+jo%2C+kala%2C+kalama%2C+kama%2C+kasi%2C+ken%2C+kepeken%2C+kili%2C+kin%2C+kipisi%2C+kiwen%2C+ko%2C+kon%2C+kule%2C+kulupu%2C+kute%2C+la%2C+lape%2C+laso%2C+lawa%2C+leko%2C+len%2C+lete%2C+li%2C+lili%2C+linja%2C+lipu%2C+loje%2C+lon%2C+luka%2C+lukin%2C+lupa%2C+ma%2C+mama%2C+mani%2C+meli%2C+mi%2C+mije%2C+moku%2C+moli%2C+monsi%2C+monsuta%2C+mu%2C+mun%2C+musi%2C+mute%2C+namako%2C+nasa%2C+nasin%2C+nena%2C+ni%2C+nimi%2C+noka%2C+o%2C+oko%2C+olin%2C+ona%2C+open%2C+pakala%2C+pali%2C+palisa%2C+pan%2C+pana%2C+pi%2C+pilin%2C+pimeja%2C+pini%2C+pipi%2C+poka%2C+poki%2C+pona%2C+pu%2C+sama%2C+seli%2C+selo%2C+seme%2C+sewi%2C+sijelo%2C+sike%2C+sin%2C+sina%2C+sinpin%2C+sitelen%2C+sona%2C+soweli%2C+suli%2C+suno%2C+supa%2C+suwi%2C+tan%2C+taso%2C+tawa%2C+telo%2C+tenpo%2C+toki%2C+tomo%2C+tu%2C+unpa%2C+uta%2C+utala%2C+walo%2C+wan%2C+waso%2C+wawa%2C+weka%2C+wile%2C+‌nanpa&minSentLen=1&scale=rel&start=1501545600&end=1722470400&smoothing=2&smoother=cwin

[quaint wagon]

tan.... seme

mi ken LUKIN A e ni: ILO LI PILIN IKE TAN MUTE NIMI

[hidden tide]

i'll just post screenshots

first off, the absolute scale! you can see how toki pona has grown over time

(i am interested to know what occurred the first half of 2023 where there is a notable trough)

[icy turret]

watch the sptp2024 ilo muni presentation!

[hidden tide]

it's more visible in the absolute entropy scale

[quaint wagon]

first recorded usage of the entropy scale

[hidden tide]

the entropy scale is very fun

[quaint wagon]

kulupu lawa pi ma ni (en mi) li pini e tomo ale ni:
#toki-moku #toki-nanpa #kalama-tpt #nasin-tpt #tpt-ale-kin

mi pini e ona tan ni: mi lukin e toki mute pi nasin ni:
1: "toki!"
2: "toki, sina pona ala pona?"
1: "mi pona. sina seme"
ni la toki li kama pini. jan li toki wawa ala li toki lon weka tenpo la ona li kama ala toki suli.
mi pini e tomo la mi pilin e ni: tenpo toki li kama lili, tan ni: jan li kama lukin e toki pi jan ante lon tenpo pona!

taso! mi kama sona e ni: lon la jan li wile e ijo toki. ni li kama e toki wawa mute. mi weka e tomo pi ijo toki la jan o toki lon seme? ala a. ni la ona li kama toki ala.

(sina ken ala lukin e tomo la ona ale li pini. o kama jo e poki pi lukin pini.)

a. mi pana ike e ona tu. taso, tomo pi nasin sama li lon tenpo pi weka tomo li pini.

[hidden tide]

sona!

[quaint wagon]

sina wile pali e tomo toki e kulupu la, o pana e ijo toki!

[hidden tide]

ijo toki li pona

[hidden tide]

sitelen sama, taso linja li supa ala
(nena suli loje li nimi "nena", nena suli walo li nimi "luka", nena suli laso li nimi "unpa")

[quaint wagon]

nimi nanpa li kama suli nasa tan ni: tenpo suno tu tu la jan wan li nasin e nanpa mute lon tomo wan lon ma wan.

wan. tu. tu wan. tu tu. luka. luka wan. luka tu.
ona li ni li pini lon poka pi mute ale. (ni li nanpa).

[hidden tide]

nanpa mute

sitelen ni la, sina ken lukin e suli pi kama suli nimi lon tenpo. linja walo sewi li nimi "pu." jan li toki mute ala e nimi ni la ona li sewi tawa nanpa wan.

mi supa ale e sitelen nasa la sina ken lukin e kama suli pi toki pona lon tenpo

[meager steeple]

nasin a

[quaint wagon]

nasa, toki pini pi wile sona li kama suli a lon tenpo

[devout jetty]

tenpo wan la mi lukin e toki pi tenpo weka li kama sona e ni: sike pini la toki "anu seme" li nasa lili tawa ijo Osuka. ni li nasa tawa mi! toki ni li suli a tawa nasin mi toki

toki ona la ona li kepeken toki "anu seme" lon tenpo pi mute lili taso li pilin e ni: toki "X ala X" li pona nanpa wan

[serene hollow]

oh my god 😭

[fresh sentinel]

is this when kijetesantakalu was coined?

[serene hollow]

thats the record for the usage of "suwi" from august of 2001 to aug. 2024

[fresh sentinel]

the hovered spike I mean

[icy turret]

feb 2009, too early

[devout jetty]

what happened 😭

ilo musi Anjelita ↩️

[Reply to:](#1272180068721889290 message) oh my god 😭 📎

I don't think there's much data for that time period, so maybe someone used suwi a bunch and it became a significant percentage of all toki pona recorded that month

[serene hollow]

oh

[serene hollow]

its gonna be someone who uses kawaii on a minutely basis watch