emoji-award-marseyspitmoji
Unable to load image

actually interesting rDrama Word Vectors

https://github.com/float-trip/wordvectors, https://github.com/float-trip/wordvectors/archive/refs/heads/master.zip

The easiest way to use these is through the included kit.py or projector.tensorflow.org.

kit.py can do the king - crown = man vector arithmetic, and the projector makes the words go spinny:

(Go to https://projector.tensorflow.org -> "Load" -> Select vectors.tsv and metadata.tsv -> Add search query, click label -> "Isolate Points")

Lots of fun things you can do with these, like finding the Marsey you're most closely associated with or clustering everyone into houses.

37
Jump in the discussion.

No email address required.

>clustering everyone into houses

house 4 chads stay winning

:#marseygigachad:

Jump in the discussion.

No email address required.

https://i.rdrama.net/images/17006887291796334.webp

Jump in the discussion.

No email address required.

House 5 GODS, keep rocking

:marseyheavymetal:

Jump in the discussion.

No email address required.

Hello House 5 friend :marseychristmasgift:

:marseycapy::marseycarp::marseyonemediumsnekvibing: :marseybride::marseyidio3::marseyklenny3::marseylaptopkrayon2:

Jump in the discussion.

No email address required.

you forgot our ace in the hole

:#marseykrayon:

Jump in the discussion.

No email address required.

How could I forget?

Jump in the discussion.

No email address required.

Heck's yeah dude! House 4 for lyfe! :marseysunglasseson:

Jump in the discussion.

No email address required.

!codecels this is way cool

unfortunately I am stuck in dependency HELL trying to get spacy to work, if anyone knows how to solve

ModuleNotFoundError: No module named 'spacy.lang.en.tag_map'

let me know 😅 (yes I ran python -m spacy download en_core_web_sm and it DIDN'T WORK :marseyraging: )

Jump in the discussion.

No email address required.

magnitude needs python 3.9 (asdf install python 3.9.12 && asdf local python 3.9.12 if you have asdf)

then it should just be pip install pymagnitude matplotlib networkx numpy scikit-learn

Jump in the discussion.

No email address required.

I got it with a venv :marseythumbsup: now where is frequencies.json? I don't see it in the repo, I guess it contains how often each token appears?

Jump in the discussion.

No email address required.

Any reason for not using gensim's word2vec. This sparse representation seems off. Anyway, great job.

I love your verification checks by using drama related analogies. Very fetch.

Jump in the discussion.

No email address required.

i cant parse this information

>finding the Marsey you're most closely associated with

can you do this for me?

Jump in the discussion.

No email address required.

here's everyone's. you had the username Madelyn at the time (I thought I fixed all those :marseygunshotsuicide:) so it might've gotten confused by that, idk if these make any sense to you

[(':indiandoomergirl:', 0.87592167), (':whitewomenfuckdogscoomer:', 0.8701779), (':holdupjak:', 0.86970484), (':hewillnever:', 0.8655615), (':sadwomanjak:', 0.8646576)]

Jump in the discussion.

No email address required.

!followers check this shit out

Jump in the discussion.

No email address required.

HAHAHAHA

(':rukiddingme:', 0.8824042), (':stfupussy:', 0.87889135), (':alienfinger:', 0.8780602), (':itschiover:', 0.8708701), (':froth:', 0.8695579)]

Jump in the discussion.

No email address required.

johnnybutt [(':dudeweedlmao:', 0.8167942), (':rukiddingme:', 0.81444925), (':itschiover:', 0.8097293), (':shakefist:', 0.80772984), (':hewillnever:', 0.8033208)]

Jump in the discussion.

No email address required.

lmao

@X [(':chiobu:', 0.89580166), (':marseypasschingchong:', 0.8835933), (':marseysingapore:', 0.88165927), (':marseybackingintobush:', 0.8781101), (':marseychingchongpearlclutch:', 0.87496513)]

i've barely used :chiobu: tho :marseysad:

Jump in the discussion.

No email address required.

it's based off the context in which your username appears, so half of the data is built from mentions by others

Jump in the discussion.

No email address required.

so half of the data is built from mentions by others

so many cyber-bullies :marseyill:

Jump in the discussion.

No email address required.

@Seppocide [(':marseyshitlover:', 0.8918634), (':marseyxesright:', 0.8907285), (':marseyyesandnosuperspeed:', 0.890585), (':marseyitneverbegan:', 0.886626), (':marseythinkinghallowseve:', 0.88611376)]

This makes no sense tbh

Jump in the discussion.

No email address required.

Can you give a two sentence explanation about how the associations are made? Are they used by me, used by people talking about me, or commented in posts by me? I'm really feeling too smallbrained to figure this out or what those emojis have to do with me.


https://i.rdrama.net/images/17092367509484937.webp https://i.rdrama.net/images/17093267613293715.webp https://i.rdrama.net/images/17177781034384797.webp

Jump in the discussion.

No email address required.

It sees the text like this:

https://i.rdrama.net/images/1700710441190946.webp

And produces embeddings based on the context in which words appear. I prepend usernames to comments so it knows who's writing them.

The embeddings are vectors with 123 dimensions/numbers which point out to some semantic meaning. The model learns how these relate to each other, which lets it do calculations like King - Man + Woman = Queen or Japan - Sushi + Germany = Bratwurst.

Your emojis probably don't mean anything, just that the model didn't really correlate any with you

(just watched venture bros btw, great show lotta heart :marseythumbsup:)

Jump in the discussion.

No email address required.

Do me!

@PalestinePhallUS love sucking peepee

Jump in the discussion.

No email address required.

[(':downdonger:', 0.9293713), (':upsoren:', 0.9263264), (':soren:', 0.9179058), (':@schizocelpat:', 0.9057671), (':capygitcommit:', 0.8905247)]

Most similar users: [('@PalestinePhallUS', 1.0), ('@G', 0.98456246), ('@PoloniumDonGER', 0.98300934), ('@sorensdonger', 0.968381), ('@sorenoell', 0.9345814)]

You are 54.92318% similar to keyed

You are 44.1374% similar to locked

Jump in the discussion.

No email address required.

Thank you! :marseyfingerhearts:

What does keyed and locked mean?

@PalestinePhallUS love sucking peepee.

@sorenoell

@Soren explain :marseythonk: When have we assimilated again

Jump in the discussion.

No email address required.

Do you have user similarity info for other users? :marseyvenn6:

Jump in the discussion.

No email address required.

Jump in the discussion.

No email address required.

Thanks! :marseyheart:

Jump in the discussion.

No email address required.

you had the username Madelyn at the time

When was the data from?

Jump in the discussion.

No email address required.

it's four months old, I think from the week the foid award came out

Jump in the discussion.

No email address required.

:#indiandoomergirl: gang rise up

Jump in the discussion.

No email address required.

@fartneighbor [(':idontknow:', 0.8762771), (':thistbh:', 0.8686737), (':stfupussy:', 0.8664979), (':rukiddingme:', 0.86619896), (':hesrightyouknow:', 0.8643853)]

Jump in the discussion.

No email address required.

>you had the username Madelyn at the time

:marseyeggirl:

and thats the tea, sis

>[(':indiandoomergirl:', 0.87592167), (':whitewomenfuckdogscoomer:', 0.8701779), (':holdupjak:', 0.86970484), (':hewillnever:', 0.8655615), (':sadwomanjak:', 0.8646576)]

yeah, these make no sense to me

(besides the womenfrickdogscoomer :marseywink:)

Jump in the discussion.

No email address required.

Jump in the discussion.

No email address required.

Fun fact: This was the first AI generated Marsey to be accepted, so very sentient considering who float trip is. It was also my first Marsey :marseyembrace:

Jump in the discussion.

No email address required.

8990737)]

@YakubianPhysics [(':iseewhatyoudidthere:', 0.8873305), (':kylieface:', 0.8826114), (':stfu:', 0.8775532), (':uzi:', 0.87725824), (':dasrite:', 0.87722605)]

lol I love it

Jump in the discussion.

No email address required.

Link copied to clipboard
Action successful!
Error, please refresh the page and try again.