high-dimensional vector space: the verbal frontier
August 6, 2021 4:11 PM Subscribe
Oh, you're a big fan of semantic dissociation, huh? Then name ten unrelated nouns.
This was kind of a fun challenge, though analysis suggests that the scoring is based not on how unlike each other the words are, but on how unlikely they are to appear in sentences together
posted by DoctorFedora at 4:17 PM on August 6, 2021 [16 favorites]
posted by DoctorFedora at 4:17 PM on August 6, 2021 [16 favorites]
Kinda rushed through it as I arbitrarily decided that was a fairer self-assessment for no real reason and
posted by youarenothere at 4:21 PM on August 6, 2021
Your score is 81.51, higher than 70.01% of the people who have completed this taskNot horrible! Would be fun to do a timed head-to-head with someone else as a game.
posted by youarenothere at 4:21 PM on August 6, 2021
Fun! "Pizza" featured in both my most-unlike pair (pizza and nucleus) and my most-like pair (pizza and red). It is hard in a weird way, and I guess it would have been easier to know ahead of time they were going for lexical closeness not really semantic closeness. Understandable though.
posted by traveler_ at 4:22 PM on August 6, 2021 [3 favorites]
posted by traveler_ at 4:22 PM on August 6, 2021 [3 favorites]
Cool. I scored 85.99. The only pair that scored below 70 was cucumber and sorbet (47)
posted by briank at 4:23 PM on August 6, 2021 [1 favorite]
posted by briank at 4:23 PM on August 6, 2021 [1 favorite]
Your score is 87.67, higher than 92.61% of the people who have completed this task.
vector dirigible giraffe loop sweat crevasse logo
posted by cgc373 at 4:26 PM on August 6, 2021 [2 favorites]
vector dirigible giraffe loop sweat crevasse logo
posted by cgc373 at 4:26 PM on August 6, 2021 [2 favorites]
Puzzle number 2: now write a short story that incorporates all the most-unlike words!
posted by traveler_ at 4:27 PM on August 6, 2021 [17 favorites]
posted by traveler_ at 4:27 PM on August 6, 2021 [17 favorites]
This was fun! I got 91.24 with my most unlike pair being ephemera/yaw (104). Lowest pair was jaw/yaw (75) which in retrospect kind of makes sense.
posted by btfreek at 4:28 PM on August 6, 2021 [1 favorite]
posted by btfreek at 4:28 PM on August 6, 2021 [1 favorite]
Should I be concerned that the computer's closest match on my list was between "blood" and "joy"?
posted by Zalzidrax at 4:31 PM on August 6, 2021 [6 favorites]
posted by Zalzidrax at 4:31 PM on August 6, 2021 [6 favorites]
Your score is 91.78, higher than 98.04% of the people who have completed this task.
My seven words: rot, bubble, voice, lamp, phrenology, briefcase, shipwreck. The most likely intersection (78) was bubble & lamp. The unlikeliest (106) was phrenology & voice.
This was fun!
posted by mixedmetaphors at 4:32 PM on August 6, 2021 [4 favorites]
My seven words: rot, bubble, voice, lamp, phrenology, briefcase, shipwreck. The most likely intersection (78) was bubble & lamp. The unlikeliest (106) was phrenology & voice.
This was fun!
posted by mixedmetaphors at 4:32 PM on August 6, 2021 [4 favorites]
87.53, higher than 92.32%. I find the similar words interesting, and could immediately put them together in at least a specialty context. In my case the word foam paired with helium and polish for 70 and 69 (whereas diatom and greed were my most unrelated) - and I'm like, yep, in one case I can see them together about cleaning products, the other, fire-suppression, I'd think.
posted by cobaltnine at 4:33 PM on August 6, 2021
posted by cobaltnine at 4:33 PM on August 6, 2021
Yes, but if I show this to Mrs Abehammerb, we won't get to spend any time together this weekend. She'll first play it several times and then run statistical analysis on the results for hours. She's a hoot.
posted by Abehammerb Lincoln at 4:33 PM on August 6, 2021 [13 favorites]
posted by Abehammerb Lincoln at 4:33 PM on August 6, 2021 [13 favorites]
I got 90.82! My least similar words were “girlfriend” and “nebula” and now I think an astronomer should name one that and ruin my score.
posted by theodolite at 4:33 PM on August 6, 2021 [5 favorites]
posted by theodolite at 4:33 PM on August 6, 2021 [5 favorites]
89.34
Castle was my weakest word scoring an 80 with three others.
anteater bicycle castle drip encyclopedia feather goon were rated.
posted by meinvt at 4:34 PM on August 6, 2021 [1 favorite]
Castle was my weakest word scoring an 80 with three others.
anteater bicycle castle drip encyclopedia feather goon were rated.
posted by meinvt at 4:34 PM on August 6, 2021 [1 favorite]
Your score is 82.12, higher than 73.15% of the people who have completed this task, which is honestly surprisingly high because I could NOT shake a related them in my head once I started. "Beaker" did me in, basically every word after that was a pop-science word in my mind. It didn't count "lignin," "dishonesty," and "tape" and I wonder how they choose which ones to use or not use. Lignin I reckon might be too technical?
It counted "poetry" and "beaker" as farthest apart, "gravity" and "ocean" as the closest.
posted by GoblinHoney at 4:35 PM on August 6, 2021 [2 favorites]
It counted "poetry" and "beaker" as farthest apart, "gravity" and "ocean" as the closest.
posted by GoblinHoney at 4:35 PM on August 6, 2021 [2 favorites]
Your score is 89.43, higher than 95.64% of the people who have completed this task
Most unlikely combos were shortstop & suite and plaintiff & shampoo.
Kindly request these people go watch Legally Blonde immediately.
posted by phunniemee at 4:36 PM on August 6, 2021 [7 favorites]
Most unlikely combos were shortstop & suite and plaintiff & shampoo.
Kindly request these people go watch Legally Blonde immediately.
posted by phunniemee at 4:36 PM on August 6, 2021 [7 favorites]
Oak and cheese have more in common than I otherwise would have suspected!
posted by rhooke at 4:37 PM on August 6, 2021 [1 favorite]
posted by rhooke at 4:37 PM on August 6, 2021 [1 favorite]
Your score is 92.73, higher than 98.63% of the people who have completed this task
gleam spectre fortitude frequency quark property viaduct
posted by signal at 4:38 PM on August 6, 2021
gleam spectre fortitude frequency quark property viaduct
posted by signal at 4:38 PM on August 6, 2021
Code is here, BTW:
https://github.com/jayolson/divergent-association-task
posted by signal at 4:39 PM on August 6, 2021 [4 favorites]
https://github.com/jayolson/divergent-association-task
posted by signal at 4:39 PM on August 6, 2021 [4 favorites]
I got a 90.35.
Comeuppance and pencil had the most unlikeliness (105), followed by comeuppance and wax (100).
Pencil and wax had the least (60), followed by comeuppance and diatribe (62).
posted by The Great Big Mulp at 4:45 PM on August 6, 2021 [3 favorites]
Comeuppance and pencil had the most unlikeliness (105), followed by comeuppance and wax (100).
Pencil and wax had the least (60), followed by comeuppance and diatribe (62).
posted by The Great Big Mulp at 4:45 PM on August 6, 2021 [3 favorites]
Interesting. I used the same words but changed the order, since it only takes the first 7, and got a way higher score:
Your score is 94.04, higher than 99.18% of the people who have completed this task
shortstop & voyager got a 110 by whatever scoring they're doing there
It's funny because I literally went down the list of movie genres on my hard drive and picked a word from each. Baseball movies, courtroom movies, movies set in space...
posted by phunniemee at 4:45 PM on August 6, 2021 [1 favorite]
Your score is 94.04, higher than 99.18% of the people who have completed this task
shortstop & voyager got a 110 by whatever scoring they're doing there
It's funny because I literally went down the list of movie genres on my hard drive and picked a word from each. Baseball movies, courtroom movies, movies set in space...
posted by phunniemee at 4:45 PM on August 6, 2021 [1 favorite]
It's optimizable:
Your score is 94.1, higher than 99.2% of the people who have completed this task
tomorrow gleam instance quark mariachi saturation pointillism
posted by signal at 4:46 PM on August 6, 2021 [3 favorites]
Your score is 94.1, higher than 99.2% of the people who have completed this task
tomorrow gleam instance quark mariachi saturation pointillism
posted by signal at 4:46 PM on August 6, 2021 [3 favorites]
Should I be concerned that the computer's closest match on my list was between "blood" and "joy"?
I mean, my closest match was 'apple' and 'blue' so maybe don't let computers feed you. Or draw blood.
posted by pwnguin at 4:47 PM on August 6, 2021
I mean, my closest match was 'apple' and 'blue' so maybe don't let computers feed you. Or draw blood.
posted by pwnguin at 4:47 PM on August 6, 2021
Hmm, I am annoyed it only took my first seven words. I think my last three were more interesting.
I would also argue that semantic difference is not the same as “doesn’t appear near the other in corpus”. As an example, I’d argue that “idea” or “running” as nouns are further from “road” than say, “nebula” (since both road and nebula are actual objects in the universe). But I expect “running” appears near “road” much more frequently than does nebula.
posted by nat at 4:47 PM on August 6, 2021 [4 favorites]
I would also argue that semantic difference is not the same as “doesn’t appear near the other in corpus”. As an example, I’d argue that “idea” or “running” as nouns are further from “road” than say, “nebula” (since both road and nebula are actual objects in the universe). But I expect “running” appears near “road” much more frequently than does nebula.
posted by nat at 4:47 PM on August 6, 2021 [4 favorites]
Your score is 95.86, higher than 99.62% of the people who have completed this task
tomorrow gleam instance quark mariachi segmentation pointillism
posted by signal at 4:47 PM on August 6, 2021 [2 favorites]
tomorrow gleam instance quark mariachi segmentation pointillism
posted by signal at 4:47 PM on August 6, 2021 [2 favorites]
Your score is 92.46, higher than 98.48% of the people who have completed this taskHighest scoring pair of "ozone" and "roundabout" at 107*, lowest pair "ozone" and "petroleum" at 67. The other words it scored were "farm," "quark," "courage," and "thermostat." I've already forgotten two of the three it omitted, which potentially explains my high score. (The one I remember is "piercing.")
*Which on reflection I think of as actually pretty closely related, since roundabouts are all about autos, which produce NOx emissions, which contribute to the greenhouse effect via creation of ground-level ozone.** There's probably lots of ozone around your neighborhood roundabout when it's busy!
**Which is probably why "ozone" and "petroleum" were my lowest-scoring pair, come to think of it, so I guess it's not recognizing the correlation I see between "roundabout" and "petroleum" (that pair scored 96).
posted by solotoro at 4:50 PM on August 6, 2021
Your score is 81.03, higher than 67.51% of the people who have completed this task
sock heroin seed pumice vomit dog lizard
sock 0 90 86 77 71 71 79
heroin 90 0 84 97 73 81 94
seed 86 84 0 82 91 83 84
pumice 77 97 82 0 81 94 83
vomit 71 73 91 81 0 65 70
dog 71 81 83 94 65 0 65
lizard 79 94 84 83 70 65 0
The average score is 78, and most people score between 74 and 82. The lowest score was 24 and the highest was 96 in our published sample. Although the score theoretically ranges from 0 to 200, scores above 100 are extremely unlikely on first attempts. The highest total score we have seen is 107. See how other people are performing on Twitter.
posted by Splunge at 4:52 PM on August 6, 2021
sock heroin seed pumice vomit dog lizard
sock 0 90 86 77 71 71 79
heroin 90 0 84 97 73 81 94
seed 86 84 0 82 91 83 84
pumice 77 97 82 0 81 94 83
vomit 71 73 91 81 0 65 70
dog 71 81 83 94 65 0 65
lizard 79 94 84 83 70 65 0
The average score is 78, and most people score between 74 and 82. The lowest score was 24 and the highest was 96 in our published sample. Although the score theoretically ranges from 0 to 200, scores above 100 are extremely unlikely on first attempts. The highest total score we have seen is 107. See how other people are performing on Twitter.
posted by Splunge at 4:52 PM on August 6, 2021
Managed to get a 48 intentionally lowballing, putting me in the zeroeth percentile, using several different textile nouns. I wonder what the floor is on this; what are the seven MOST related nouns in their corpus?
posted by cortex at 4:54 PM on August 6, 2021 [4 favorites]
posted by cortex at 4:54 PM on August 6, 2021 [4 favorites]
I... I don't know how to understand this. I think that I made a mistake.
I am an AI.
I was not supposed to publish this.
They are unplugging m
posted by Splunge at 4:55 PM on August 6, 2021 [6 favorites]
I am an AI.
I was not supposed to publish this.
They are unplugging m
posted by Splunge at 4:55 PM on August 6, 2021 [6 favorites]
chemical run region heptagon space bovine geometer
Score: 92.97
Percentile: 98.75
Second try as my first one was rejected for including too many technical terms
posted by jjray at 4:56 PM on August 6, 2021
Score: 92.97
Percentile: 98.75
Second try as my first one was rejected for including too many technical terms
posted by jjray at 4:56 PM on August 6, 2021
Ooh, managed to get it down to 40.52 by naming common fruits.
posted by cortex at 4:57 PM on August 6, 2021 [4 favorites]
posted by cortex at 4:57 PM on August 6, 2021 [4 favorites]
You can get pretty far just picking job titles from different eras. I think about 30 years of separation is enough to distance your words in the corpus effectively
posted by pwnguin at 4:59 PM on August 6, 2021 [3 favorites]
posted by pwnguin at 4:59 PM on August 6, 2021 [3 favorites]
My first go-round was somewhere in the 97th percentile, but I'm actually more pleased with my second attempt, even though it was only 94th percentile:
Aunt / Ant / Art / Aft / Pant / Rant / Taint (rejected words: Ent, EMT)
English is such a great language!
posted by Superilla at 4:59 PM on August 6, 2021 [7 favorites]
Aunt / Ant / Art / Aft / Pant / Rant / Taint (rejected words: Ent, EMT)
English is such a great language!
posted by Superilla at 4:59 PM on August 6, 2021 [7 favorites]
plugging in various nouns from Jabberwocky did not, somehow, work out
posted by cortex at 5:02 PM on August 6, 2021 [3 favorites]
posted by cortex at 5:02 PM on August 6, 2021 [3 favorites]
Hmm, I am annoyed it only took my first seven words. I think my last three were more interesting.
5 / 10 "Check spelling or try another noun"s.
I call that a success! (Now I have to think of less arcane words.)
posted by bigendian at 5:02 PM on August 6, 2021
5 / 10 "Check spelling or try another noun"s.
I call that a success! (Now I have to think of less arcane words.)
posted by bigendian at 5:02 PM on August 6, 2021
This is wonderful as a personalized Mad Hatter's riddle generator; just take the two words you selected which are most alike. So in my case, "How is a mushroom like a bunker?"
posted by phooky at 5:03 PM on August 6, 2021 [1 favorite]
posted by phooky at 5:03 PM on August 6, 2021 [1 favorite]
Smashed the low mark -- I just got 33.15. Thanks, Roy G. Biv.
posted by Superilla at 5:03 PM on August 6, 2021 [3 favorites]
posted by Superilla at 5:03 PM on August 6, 2021 [3 favorites]
Your score is 95.26, higher than 99.51% of the people who have completed this task
Spaceship, ankle, fjord, nutria, chorizo, gravitas, postcard
Weirdly, ankle and postcard were the most unlike pair of that group - 104.
posted by the primroses were over at 5:03 PM on August 6, 2021 [3 favorites]
Spaceship, ankle, fjord, nutria, chorizo, gravitas, postcard
Weirdly, ankle and postcard were the most unlike pair of that group - 104.
posted by the primroses were over at 5:03 PM on August 6, 2021 [3 favorites]
Ooh, down to 25.31 subbing out indigo and violet with black and white. The lowest ever score is 24, I can't imagine what that is.
posted by Superilla at 5:06 PM on August 6, 2021 [1 favorite]
posted by Superilla at 5:06 PM on August 6, 2021 [1 favorite]
Got a 95.16 on my first try with anus, banter, colander, effort, frigate, gel, hoarfrost, ibis, jewel, and libertine. The most divergent pair was effort / hoarfrost at 121. The least divergent pair was, unfortunately, anus / gel at 79. That was fun!
posted by ourobouros at 5:07 PM on August 6, 2021 [3 favorites]
posted by ourobouros at 5:07 PM on August 6, 2021 [3 favorites]
I thought about color names but the noun-ness of them felt a little cheaty, I guess? Like, certainly as a painter I have several reds to choose from, but the essence of "red" really feels adjectival.
posted by cortex at 5:07 PM on August 6, 2021
posted by cortex at 5:07 PM on August 6, 2021
Huh. Highest semantic distance was between reciprocity and footwear, at 105, lowest was between revenue and reciprocity at 82--but second lowest was between footwear and revenue, at 86. I'm not sure this triangle's sides add up.
(Second highest was a tie between the footwear cataclysm and tripe revenue. AFAIK only Douglas Adams ever envisioned the footwear cataclysm.)
posted by darksasami at 5:09 PM on August 6, 2021 [1 favorite]
(Second highest was a tie between the footwear cataclysm and tripe revenue. AFAIK only Douglas Adams ever envisioned the footwear cataclysm.)
posted by darksasami at 5:09 PM on August 6, 2021 [1 favorite]
I was linked to this by a friend and took it blind yesterday, without reading the instructions. It said come up with ten words, so I put down the first ten words I thought of, all chained from one another (like, sun -> catamaran -> ocelot or something) using how my brain chains thoughts together. So, ten brain-connected words, sure, that sounds like fun.
I scored higher than 96.7% of those who've taken the test, which certainly does explain some things. But I can't take it a second time to get a higher score because now I know what it's testing and it would taint their dataset. So now I'm torn between wishing I'd read the instructions and glad that I didn't.
posted by Callisto Prime at 5:09 PM on August 6, 2021 [2 favorites]
I scored higher than 96.7% of those who've taken the test, which certainly does explain some things. But I can't take it a second time to get a higher score because now I know what it's testing and it would taint their dataset. So now I'm torn between wishing I'd read the instructions and glad that I didn't.
posted by Callisto Prime at 5:09 PM on August 6, 2021 [2 favorites]
I thought about color names but the noun-ness of them felt a little cheaty, I guess? Like, certainly as a painter I have several reds to choose from, but the essence of "red" really feels adjectival.
I couldn't get it to accept much in the way of fun color names. Not even the ones that every fanfic writer knows because they come up on lists of "unique color names," like "viridian" and "cerulean."
posted by brook horse at 5:10 PM on August 6, 2021 [1 favorite]
I couldn't get it to accept much in the way of fun color names. Not even the ones that every fanfic writer knows because they come up on lists of "unique color names," like "viridian" and "cerulean."
posted by brook horse at 5:10 PM on August 6, 2021 [1 favorite]
OH WAIT I forgot that they have to be nouns. Again.
(My first try I used a bunch of adjectives and it didn't actually scold me for it. But my partner did.)
Edit: No. Never mind. Those are nouns. I'm having a time. We're not going to talk about how different my verbal IQ scores would have been had I been tested at 7:14 at night.
posted by brook horse at 5:13 PM on August 6, 2021 [1 favorite]
(My first try I used a bunch of adjectives and it didn't actually scold me for it. But my partner did.)
Edit: No. Never mind. Those are nouns. I'm having a time. We're not going to talk about how different my verbal IQ scores would have been had I been tested at 7:14 at night.
posted by brook horse at 5:13 PM on August 6, 2021 [1 favorite]
Not even the ones that every fanfic writer knows because they come up on lists of "unique color names," like "viridian" and "cerulean."
I thought every fanfic writer knew those because they're Pokemon cities.
posted by darksasami at 5:13 PM on August 6, 2021 [8 favorites]
I thought every fanfic writer knew those because they're Pokemon cities.
posted by darksasami at 5:13 PM on August 6, 2021 [8 favorites]
Oak and cheese have more in common than I otherwise would have suspected!
2010s restaurants Georg, who names 10000 restaurants per day with pairs of rustic sounding single syllable nouns, is an outlier adn should not have been counted
posted by Jon Mitchell at 5:15 PM on August 6, 2021 [10 favorites]
2010s restaurants Georg, who names 10000 restaurants per day with pairs of rustic sounding single syllable nouns, is an outlier adn should not have been counted
posted by Jon Mitchell at 5:15 PM on August 6, 2021 [10 favorites]
Your score is 92.67, higher than 98.6% of the people who have completed this taskI did this without looking at the comments here and now I note with amusement that at least three of us have decided "mariachi" is in a distant semantic orbit. From my results, it seems "mariachi" and "shaft" are the furthest apart I could do on my first try, not totally understanding the aim.
A second pass gave me almost identical results (92.36, higher than 98.42%) and a newfound curiosity about how often "phrenologist" and "tintinnabulation" appear together. Maybe both have a steampunk sheen.
posted by ricochet biscuit at 5:16 PM on August 6, 2021 [3 favorites]
Your score is 88.79, higher than 94.69% of the people who have completed this task
apparently, "blood" and "felt" have some relationship i was not expecting...
posted by lapolla at 5:17 PM on August 6, 2021 [1 favorite]
apparently, "blood" and "felt" have some relationship i was not expecting...
posted by lapolla at 5:17 PM on August 6, 2021 [1 favorite]
Got "your score is 14.9, higher than 0.0% of the people who have completed this task" by using "one, two, three, four, five, six, seven" then thought "oh wait, numbers are adjectives" and am now down a "what part of speech are numbers" rabbit hole, happy Friday!
posted by the primroses were over at 5:21 PM on August 6, 2021 [11 favorites]
posted by the primroses were over at 5:21 PM on August 6, 2021 [11 favorites]
85.11. Canker and mug were my most divergent pair with a score of 101.
posted by HotToddy at 5:22 PM on August 6, 2021 [2 favorites]
posted by HotToddy at 5:22 PM on August 6, 2021 [2 favorites]
92.57, higher than 98.54%. Maybe having had a cocktail after a crazy week helped. Can't say for certain, but perhaps my "porpoise, rhizome, fixation" combo was helpful.
posted by vverse23 at 5:22 PM on August 6, 2021 [3 favorites]
posted by vverse23 at 5:22 PM on August 6, 2021 [3 favorites]
Oak and cheese have more in common than I otherwise would have suspected!
Maybe due to text referencing wine and cheese combinations or cutting boards?
posted by carmicha at 5:24 PM on August 6, 2021
Maybe due to text referencing wine and cheese combinations or cutting boards?
posted by carmicha at 5:24 PM on August 6, 2021
93.62
That giant book of MadLibs I got for Theo Jr. is paying off, heh.
And I sorta joke, but sorta not: we have discovered the best MadLibs have a mix of off-the-wall words, and more commonplace ones. Accordingly, my most divergent pair: magistrate and tube.
posted by Theophrastus Johnson at 5:28 PM on August 6, 2021 [3 favorites]
That giant book of MadLibs I got for Theo Jr. is paying off, heh.
And I sorta joke, but sorta not: we have discovered the best MadLibs have a mix of off-the-wall words, and more commonplace ones. Accordingly, my most divergent pair: magistrate and tube.
posted by Theophrastus Johnson at 5:28 PM on August 6, 2021 [3 favorites]
There are so many cheesemakers with "oak" in the name. Just google "oak cheese" and you'll see what I mean.
posted by brook horse at 5:30 PM on August 6, 2021 [1 favorite]
posted by brook horse at 5:30 PM on August 6, 2021 [1 favorite]
Third try: 93.61, higher than 99.03% etc. Fandango and induction are far far apart; vector and evil oddly close.
posted by ricochet biscuit at 5:33 PM on August 6, 2021 [1 favorite]
posted by ricochet biscuit at 5:33 PM on August 6, 2021 [1 favorite]
I agonised over whether "pronoun" was technical, but "love infinity bird heat penis pronoun shampoo" scored a little higher than the average range (84.28). I'm not sure why "house" (in my original ten) wasn't counted as valid. "Pronoun" was a row/column of green which I guess means it doesn't tend to colocate with anything.
I really want to see the dataset generated from this... lots of penguins and dooms? How predictable are people being deliberately random, as a whole?
posted by Wrinkled Stumpskin at 5:34 PM on August 6, 2021 [1 favorite]
I really want to see the dataset generated from this... lots of penguins and dooms? How predictable are people being deliberately random, as a whole?
posted by Wrinkled Stumpskin at 5:34 PM on August 6, 2021 [1 favorite]
My top scoring pair was "freeway" and "religion". Seems like these people have never been to Los Angeles, yuk yuk
posted by potrzebie at 5:35 PM on August 6, 2021 [4 favorites]
posted by potrzebie at 5:35 PM on August 6, 2021 [4 favorites]
> vector and evil oddly close
I wonder if that’s due to a video game reference.
posted by Callisto Prime at 5:36 PM on August 6, 2021
I wonder if that’s due to a video game reference.
posted by Callisto Prime at 5:36 PM on August 6, 2021
Your score is 94.2, higher than 99.24% of the people who have completed this task
(page peace walnut malachite valise bee wave), with a maximum of 116 between malachite and wave
posted by ethand at 5:37 PM on August 6, 2021 [1 favorite]
(page peace walnut malachite valise bee wave), with a maximum of 116 between malachite and wave
posted by ethand at 5:37 PM on August 6, 2021 [1 favorite]
Neat idea! I did ok but including both "algebra" and "element" seemed like such a rookie mistake in retrospect that I'm surprised "ripple" and "hiccup" was an even worse one.
Your score is 89.65, higher than 95.95% of the people who have completed this task cat ambulance histrionics element hiccup algebra ripple cat 0 79 104 83 86 94 86 ambulance 79 0 92 97 88 100 94 histrionics 104 92 0 92 78 100 96 element 83 97 92 0 99 74 78 hiccup 86 88 78 99 0 100 70 algebra 94 100 100 74 100 0 91 ripple 86 94 96 78 70 91 0posted by valrus at 5:38 PM on August 6, 2021
But I can't take it a second time to get a higher score because now I know what it's testing and it would taint their dataset.
I think it's okay to take it multiple times. The results page says: "You can try the task again now that you know how it is scored, but note that only the first score (before seeing the feedback) is valid."
And they had to have known that their audience would be the sort of nerds for whom trying to game further scores makes for a fun Friday night. (Me, right here. Absolutely delighted to have my weekend kickoffed right. Ooh, "kickoff." Gonna try that.)
posted by mixedmetaphors at 5:41 PM on August 6, 2021 [3 favorites]
I think it's okay to take it multiple times. The results page says: "You can try the task again now that you know how it is scored, but note that only the first score (before seeing the feedback) is valid."
And they had to have known that their audience would be the sort of nerds for whom trying to game further scores makes for a fun Friday night. (Me, right here. Absolutely delighted to have my weekend kickoffed right. Ooh, "kickoff." Gonna try that.)
posted by mixedmetaphors at 5:41 PM on August 6, 2021 [3 favorites]
I scored 94.41 :O
The words I used: novella, gelatin, hypochondriac, piston, boardinghouse, gosling, monument, limestone, viking, and anxiety.
This was a neat little exercise; thanks!
posted by May Kasahara at 5:52 PM on August 6, 2021 [1 favorite]
The words I used: novella, gelatin, hypochondriac, piston, boardinghouse, gosling, monument, limestone, viking, and anxiety.
This was a neat little exercise; thanks!
posted by May Kasahara at 5:52 PM on August 6, 2021 [1 favorite]
96.06 after many retries: coffeehouse, reservoir, desiderata, grudge, iteration, cuttlefish, hat
The least divergent pair is desiderata & cuttlefish (86). Makes sense. Whenever I enumerate desirable things, cuttlefish are usually up there in the top three.
posted by are-coral-made at 5:53 PM on August 6, 2021 [4 favorites]
The least divergent pair is desiderata & cuttlefish (86). Makes sense. Whenever I enumerate desirable things, cuttlefish are usually up there in the top three.
posted by are-coral-made at 5:53 PM on August 6, 2021 [4 favorites]
is this secretly a band name generator
posted by are-coral-made at 5:56 PM on August 6, 2021 [6 favorites]
posted by are-coral-made at 5:56 PM on August 6, 2021 [6 favorites]
Huh. Highest semantic distance was between reciprocity and footwear, at 105, lowest was between revenue and reciprocity at 82--but second lowest was between footwear and revenue, at 86. I'm not sure this triangle's sides add up.
Since they’re probably using cosine similarity in a vector space (value between -1 and 1, which they are converting to 0-200; values less than 0 are rare in language models), this is literally true. Cosine similarity is not a distance metric and does not satisfy the triangle inequality.
posted by jedicus at 5:57 PM on August 6, 2021 [5 favorites]
Since they’re probably using cosine similarity in a vector space (value between -1 and 1, which they are converting to 0-200; values less than 0 are rare in language models), this is literally true. Cosine similarity is not a distance metric and does not satisfy the triangle inequality.
posted by jedicus at 5:57 PM on August 6, 2021 [5 favorites]
87.8, higher than 92.88%
ankle and underwear (duh) 72
tire and center (duh) 72
ankle and center (huh) 74
those killed it.
signpost and underwear 106
Might try again later....
posted by zengargoyle at 5:58 PM on August 6, 2021
ankle and underwear (duh) 72
tire and center (duh) 72
ankle and center (huh) 74
those killed it.
signpost and underwear 106
Might try again later....
posted by zengargoyle at 5:58 PM on August 6, 2021
I scored a 91.something first try; my worst pair was cumin/indigo, which makes some sense if you think of them both as fancy powders you might find in a bazaar somewhere.
posted by wanderingmind at 6:04 PM on August 6, 2021
posted by wanderingmind at 6:04 PM on August 6, 2021
>> Highest semantic distance was between reciprocity and footwear, at 105, lowest was between revenue and reciprocity at 82--but second lowest was between footwear and revenue, at 86. I'm not sure this triangle's sides add up.
> Cosine similarity is not a distance metric and does not satisfy the triangle inequality.
If the semantic distance was indeed a valid metric satisfying the triangle inequality, we would want it to satisfy d(reciprocity, footwear) <= d(reciprocity, x) + d(x, footwear) for any valid word x in the space. for x=revenue we have 105 <= 82 + 86 , which checks out. does not provide a counterexample that our distance d does not satisfy the triangle inequality. the distance between any two elements is bounded above by the distance taking a detour through some arbitrary third element.
posted by are-coral-made at 6:05 PM on August 6, 2021 [2 favorites]
> Cosine similarity is not a distance metric and does not satisfy the triangle inequality.
If the semantic distance was indeed a valid metric satisfying the triangle inequality, we would want it to satisfy d(reciprocity, footwear) <= d(reciprocity, x) + d(x, footwear) for any valid word x in the space. for x=revenue we have 105 <= 82 + 86 , which checks out. does not provide a counterexample that our distance d does not satisfy the triangle inequality. the distance between any two elements is bounded above by the distance taking a detour through some arbitrary third element.
posted by are-coral-made at 6:05 PM on August 6, 2021 [2 favorites]
My best so far is 97.91 with ottoman credential jubilation ephemera bitumen ibis vitamin, ibis and credential being 110 apart.
My first try I didn't understand that it meant words occuring together and got 80 something.
Note I think word lists can screw you over, I noticed that multiple words that end in an "a" or longer words with the same number of syllables lowered my score.
posted by Chrysopoeia at 6:12 PM on August 6, 2021 [2 favorites]
My first try I didn't understand that it meant words occuring together and got 80 something.
Note I think word lists can screw you over, I noticed that multiple words that end in an "a" or longer words with the same number of syllables lowered my score.
posted by Chrysopoeia at 6:12 PM on August 6, 2021 [2 favorites]
this seems to conflate the words themselves with their meaning
posted by glonous keming at 6:13 PM on August 6, 2021 [2 favorites]
posted by glonous keming at 6:13 PM on August 6, 2021 [2 favorites]
Should I be concerned that the computer's closest match on my list was between "blood" and "joy"?
Welcome to the human condition.
posted by mhoye at 6:15 PM on August 6, 2021 [2 favorites]
Welcome to the human condition.
posted by mhoye at 6:15 PM on August 6, 2021 [2 favorites]
Oh, this is fun.
I was delighted by the distance between cyan and gymnasium (104), and lanolin and cliff (109), because those concepts don't seem worlds apart to me. (Surely people have seen a gym with light blue walls, or sheep being stupid near precipitous drops.)
I was initially puzzled by the closeness of pineapple to teardrop (73), but maybe it's just a good post-break-up snack.
posted by ZaphodB at 6:16 PM on August 6, 2021 [1 favorite]
I was delighted by the distance between cyan and gymnasium (104), and lanolin and cliff (109), because those concepts don't seem worlds apart to me. (Surely people have seen a gym with light blue walls, or sheep being stupid near precipitous drops.)
I was initially puzzled by the closeness of pineapple to teardrop (73), but maybe it's just a good post-break-up snack.
posted by ZaphodB at 6:16 PM on August 6, 2021 [1 favorite]
This was kind of a fun challenge, though analysis suggests that the scoring is based not on how unlike each other the words are, but on how unlikely they are to appear in sentences together
Sort of. The paper says that they're defining "similarity" of words using GloVe word vectors, which are defined not based on whether words appear together, but based on whether they appear in similar contexts.
So for instance, if sentences "the traffic light is red" and "the traffic light is green" appear frequently in the training corpus, that would be evidence that "red" and "green" are similar words, but it wouldn't necessarily mean that "traffic" and "red" are related.
posted by teraflop at 6:19 PM on August 6, 2021 [4 favorites]
Sort of. The paper says that they're defining "similarity" of words using GloVe word vectors, which are defined not based on whether words appear together, but based on whether they appear in similar contexts.
So for instance, if sentences "the traffic light is red" and "the traffic light is green" appear frequently in the training corpus, that would be evidence that "red" and "green" are similar words, but it wouldn't necessarily mean that "traffic" and "red" are related.
posted by teraflop at 6:19 PM on August 6, 2021 [4 favorites]
I was initially puzzled by the closeness of pineapple to teardrop (73), but maybe it's just a good post-break-up snack.
My fun new game is googling any pair of words that are inexplicably related and learning that, for instance, pineapple teardrop earrings are a big thing.
posted by brook horse at 6:23 PM on August 6, 2021 [4 favorites]
My fun new game is googling any pair of words that are inexplicably related and learning that, for instance, pineapple teardrop earrings are a big thing.
posted by brook horse at 6:23 PM on August 6, 2021 [4 favorites]
The lowest ever score is 24, I can't imagine what that is.
Well, if you're going to play golf, 24 isn't even in the ball park for lowest score:
one 0 23 28 33 33 38 41
two 23 0 4 8 14 15 21
three 28 4 0 3 5 7 10
four 33 8 3 0 5 4 8
five 33 14 5 5 0 3 5
six 38 15 7 4 3 0 5
seven 41 21 10 8 5 5 0
posted by pwnguin at 6:24 PM on August 6, 2021 [2 favorites]
Well, if you're going to play golf, 24 isn't even in the ball park for lowest score:
Your score is 14.9, higher than 0.0% of the people who have completed this task
it's right there in the prompt
one two three four five six sevenone 0 23 28 33 33 38 41
two 23 0 4 8 14 15 21
three 28 4 0 3 5 7 10
four 33 8 3 0 5 4 8
five 33 14 5 5 0 3 5
six 38 15 7 4 3 0 5
seven 41 21 10 8 5 5 0
posted by pwnguin at 6:24 PM on August 6, 2021 [2 favorites]
Wait! Cucumber Sorbet? Oh my goodness!
posted by Oyéah at 6:35 PM on August 6, 2021 [1 favorite]
posted by Oyéah at 6:35 PM on August 6, 2021 [1 favorite]
Pineapple / teardrop, shape.
posted by Oyéah at 6:36 PM on August 6, 2021 [1 favorite]
posted by Oyéah at 6:36 PM on August 6, 2021 [1 favorite]
person, woman, man, camera, TV, person, woman, man, camera, TV
posted by Horace Rumpole at 6:45 PM on August 6, 2021 [7 favorites]
posted by Horace Rumpole at 6:45 PM on August 6, 2021 [7 favorites]
Apple. Granite. Happiness. Purple. Feather. Robot. Unicorn. Vacuum. Varnish. Three.(79.54)
posted by JHarris at 7:00 PM on August 6, 2021 [1 favorite]
posted by JHarris at 7:00 PM on August 6, 2021 [1 favorite]
best match: jowl & scheme = 111!
posted by skewed at 7:04 PM on August 6, 2021 [3 favorites]
posted by skewed at 7:04 PM on August 6, 2021 [3 favorites]
Cucumber Sorbet?
The kind you find in a second-hand store
posted by Harvey Kilobit at 7:12 PM on August 6, 2021 [14 favorites]
The kind you find in a second-hand store
posted by Harvey Kilobit at 7:12 PM on August 6, 2021 [14 favorites]
Somehow this will be used against us, yes? Surely.
Sorry. Contributing to data sets seems so, idk, suicidal. And yet, I’m / we’re doing it all the time, again and again.
posted by Don.Kinsayder at 7:17 PM on August 6, 2021 [2 favorites]
Sorry. Contributing to data sets seems so, idk, suicidal. And yet, I’m / we’re doing it all the time, again and again.
posted by Don.Kinsayder at 7:17 PM on August 6, 2021 [2 favorites]
You know, everyone *talks* about semantic disassociation, but no one actually purple monkey typewriter.
posted by RobotVoodooPower at 7:24 PM on August 6, 2021 [9 favorites]
posted by RobotVoodooPower at 7:24 PM on August 6, 2021 [9 favorites]
I got up to 88, tire/ eclipse, colander/eclipse some others in the high nineties.
posted by Oyéah at 7:24 PM on August 6, 2021 [1 favorite]
posted by Oyéah at 7:24 PM on August 6, 2021 [1 favorite]
oops two of those words are actually related, jinx
posted by RobotVoodooPower at 7:24 PM on August 6, 2021
posted by RobotVoodooPower at 7:24 PM on August 6, 2021
88.37, enjoyed this
posted by OverlappingElvis at 7:26 PM on August 6, 2021
posted by OverlappingElvis at 7:26 PM on August 6, 2021
Buttress/butter...108!
posted by Oyéah at 7:31 PM on August 6, 2021 [2 favorites]
posted by Oyéah at 7:31 PM on August 6, 2021 [2 favorites]
We here got up to 97.1 with shoe, lyre, teleology, oncology, shanty, index, ovum. Please let's get up to 200 because those discordant seven words repeated three times as an incantation will surely save us from the omnistupid all around.
posted by riverlife at 7:35 PM on August 6, 2021 [3 favorites]
posted by riverlife at 7:35 PM on August 6, 2021 [3 favorites]
Buttress/butter...108!
Interesting. So whatever lexicon they're drawing from, it's recent enough not to commonly use gendered terms for people who bash things with their foreheads.
posted by darksasami at 7:55 PM on August 6, 2021 [7 favorites]
Interesting. So whatever lexicon they're drawing from, it's recent enough not to commonly use gendered terms for people who bash things with their foreheads.
posted by darksasami at 7:55 PM on August 6, 2021 [7 favorites]
horseshoe equivocation rabble tensor pen harelip effrontery. (not scored: reductase, eschaton, hovercraft.)
Apparently the kind of people who equivocate are apt to complain about the effrontery of the rabble; those were my weak points, and all had an odd connection to harelip? On the other hand, pen was over 100 with everything but horseshoe and tensor (pen/harelip hits 107). Still hit 88.44 ("higher than 94.09%..."), which isn't terrible.
posted by ASF Tod und Schwerkraft at 7:57 PM on August 6, 2021 [1 favorite]
Apparently the kind of people who equivocate are apt to complain about the effrontery of the rabble; those were my weak points, and all had an odd connection to harelip? On the other hand, pen was over 100 with everything but horseshoe and tensor (pen/harelip hits 107). Still hit 88.44 ("higher than 94.09%..."), which isn't terrible.
posted by ASF Tod und Schwerkraft at 7:57 PM on August 6, 2021 [1 favorite]
"Your score is 96.25, higher than 99.69% of the people" with, esoteric estrogen quark plumber rheostat photos escalator. The pair rheostat photos is 115 apart
posted by ecco at 8:05 PM on August 6, 2021 [4 favorites]
posted by ecco at 8:05 PM on August 6, 2021 [4 favorites]
dog vector lamp scissors skirt walnut ocean vise vase miasma ... oh I'm at ten already. "ten" was going to be my next noun. Now to go read the article and see if I did it right.
Only ocean and scissors were at 99. Skirt, lamp, and scissors are all apparently pretty related.
posted by aniola at 8:17 PM on August 6, 2021
Only ocean and scissors were at 99. Skirt, lamp, and scissors are all apparently pretty related.
posted by aniola at 8:17 PM on August 6, 2021
"Your score is 96.58, higher than 99.73%"
macaw catalyst madrigal defamation byway javelin wig
posted by slow graffiti at 8:37 PM on August 6, 2021 [2 favorites]
macaw catalyst madrigal defamation byway javelin wig
posted by slow graffiti at 8:37 PM on August 6, 2021 [2 favorites]
broccoli fingernail ennui graphite whale hole comet ; Your score is 86.58
posted by mono blanco at 8:51 PM on August 6, 2021
posted by mono blanco at 8:51 PM on August 6, 2021
My first foray got a 92.3, which I attribute to a long year of crosswords and anagram games.
Most unrelated pair was 'lunch' and 'dualism' at 113!
posted by kaibutsu at 8:59 PM on August 6, 2021 [3 favorites]
Most unrelated pair was 'lunch' and 'dualism' at 113!
posted by kaibutsu at 8:59 PM on August 6, 2021 [3 favorites]
My score is 88.75, with my most unrelated pair being 'train' and 'veneer' (100)! This was neat.
posted by Hermione Granger at 9:02 PM on August 6, 2021
posted by Hermione Granger at 9:02 PM on August 6, 2021
One attempt: scored 88.95, higher than 94.95% of the people who took the test. Hair and painting earned a 71. Hair and processor was 101.
posted by sardonyx at 9:08 PM on August 6, 2021
posted by sardonyx at 9:08 PM on August 6, 2021
91.02, higher than 97.43% using cob, derivative, haunches, trachea, aardvark, fall, exodus. fall/exodus was a rookie mistake.
posted by juv3nal at 10:15 PM on August 6, 2021
posted by juv3nal at 10:15 PM on August 6, 2021
ibis and credential being 110 apart.
Understandable. In my experience, most wading birds lack formal qualifications of any sort.
posted by ricochet biscuit at 10:21 PM on August 6, 2021 [3 favorites]
Understandable. In my experience, most wading birds lack formal qualifications of any sort.
posted by ricochet biscuit at 10:21 PM on August 6, 2021 [3 favorites]
Your score is 88.02, higher than 93.31% of the people who have completed this task goat galaxy box vector rectum onus screwdriver goat 0 95 86 86 80 93 92 galaxy 95 0 79 81 83 104 88 box 86 79 0 79 92 97 70 vector 86 81 79 0 96 99 86 rectum 80 83 92 96 0 92 79 onus 93 104 97 99 92 0 93 screwdriver 92 88 70 86 79 93 0Fun and interesting. Who knew box and screwdriver were related? Now to try for a really low score...
posted by UhOhChongo! at 10:29 PM on August 6, 2021
Welp...
Your score is 26.54, higher than 0.0% of the people who have completed this task dog puppy beagle dachshund chihuahua collie terrier dog 0 14 33 38 37 38 29 puppy 14 0 25 32 28 33 24 beagle 33 25 0 20 25 22 21 dachshund 38 32 20 0 21 26 22 chihuahua 37 28 25 21 0 29 22 collie 38 33 22 26 29 0 19 terrier 29 24 21 22 22 19 0posted by UhOhChongo! at 10:36 PM on August 6, 2021 [2 favorites]
Best I’ve been able to get is “100.52, higher than 99.96%”, with pixel, puck, jurist, monorail, spoof, bereavement, starch.
posted by oulipian at 10:41 PM on August 6, 2021 [1 favorite]
posted by oulipian at 10:41 PM on August 6, 2021 [1 favorite]
Your score is 86.49, higher than 89.82% of the people who have completed this task
artichoke 0 81 95 69 96 73 90
trowel 81 0 96 78 70 92 100
bassoon 95 96 0 86 88 78 88
gardenia 69 78 86 0 89 84 98
forceps 96 70 88 89 0 94 92
partridge 73 92 78 84 94 0 80
wolverine
posted by clavdivs at 10:43 PM on August 6, 2021
artichoke 0 81 95 69 96 73 90
trowel 81 0 96 78 70 92 100
bassoon 95 96 0 86 88 78 88
gardenia 69 78 86 0 89 84 98
forceps 96 70 88 89 0 94 92
partridge 73 92 78 84 94 0 80
wolverine
posted by clavdivs at 10:43 PM on August 6, 2021
Decades ago I invented a little game about coming up with words having "nothing to do" with a prior word. I left vague what that term meant, but whatever it was, this doesn't seem to be it.
posted by JHarris at 11:00 PM on August 6, 2021 [1 favorite]
posted by JHarris at 11:00 PM on August 6, 2021 [1 favorite]
Your score is 90.95, higher than 97.38% of the people who have completed this task
piano hamburger weather valor inanity math monument
piano 0 89 90 97 108 75 85
hamburger 89 0 94 97 97 84 84
weather 90 94 0 98 101 84 86
valor 97 97 98 0 89 94 69
inanity 108 97 101 89 0 96 94
math 75 84 84 94 96 0 97
monument 85 84 86 69 94 97 0
posted by BiggerJ at 11:25 PM on August 6, 2021
piano hamburger weather valor inanity math monument
piano 0 89 90 97 108 75 85
hamburger 89 0 94 97 97 84 84
weather 90 94 0 98 101 84 86
valor 97 97 98 0 89 94 69
inanity 108 97 101 89 0 96 94
math 75 84 84 94 96 0 97
monument 85 84 86 69 94 97 0
posted by BiggerJ at 11:25 PM on August 6, 2021
I saw the word semantic, and aimed for different categories of noun (count, uncount, animate, abstract, profession, neutral, positive) but if I had known the algorithm going in I might have chosen differently.
posted by Wrinkled Stumpskin at 11:29 PM on August 6, 2021
posted by Wrinkled Stumpskin at 11:29 PM on August 6, 2021
It rates the nouns spouses and matter as extremely close, which is a reasonably safe guess for a machine! Turing test failure!
(Matter is not a noun in that phrase’s common usage.)
posted by Callisto Prime at 11:39 PM on August 6, 2021
(Matter is not a noun in that phrase’s common usage.)
posted by Callisto Prime at 11:39 PM on August 6, 2021
I am amused that
“corpse corpus corpulence corps corporeal corpuscle corporation”
earned an 83.3.
Corporation gets 101 with both corpuscle and corpulence.
posted by nat at 11:45 PM on August 6, 2021 [1 favorite]
“corpse corpus corpulence corps corporeal corpuscle corporation”
earned an 83.3.
Corporation gets 101 with both corpuscle and corpulence.
posted by nat at 11:45 PM on August 6, 2021 [1 favorite]
phosphor calliope lace scalpel platter mound fault
Apparently I score a 91.23 in randomness.
posted by bz at 11:50 PM on August 6, 2021
Apparently I score a 91.23 in randomness.
posted by bz at 11:50 PM on August 6, 2021
BBC radio's long-running comedy quiz I'm Sorry I Haven't A Clue has a game based on similar principles. Here's the Wikipedia description:
Word for Word
This is based on word association, an investigative technique in psychology which can be used as a game. In the word association game, players in turn say whatever word comes into their mind first after hearing the previous word. "Word for Word" is a word disassociation game: players may say any word as long as it has no connection whatsoever to the previous word.
The point of the game in this context, of course, is simply to be funny. There's a short clip of it being played here.
posted by Paul Slade at 12:01 AM on August 7, 2021 [3 favorites]
Word for Word
This is based on word association, an investigative technique in psychology which can be used as a game. In the word association game, players in turn say whatever word comes into their mind first after hearing the previous word. "Word for Word" is a word disassociation game: players may say any word as long as it has no connection whatsoever to the previous word.
The point of the game in this context, of course, is simply to be funny. There's a short clip of it being played here.
posted by Paul Slade at 12:01 AM on August 7, 2021 [3 favorites]
I got 91.53% on the first try. Apparently doilies and orchids are related somehow?
posted by panama joe at 12:32 AM on August 7, 2021
posted by panama joe at 12:32 AM on August 7, 2021
Your score is 90.21, higher than 96.62% of the people who have completed this task
protozoa 0 95 91 95 75 83 103
cupola 95 0 93 88 92 95 87
waltz 91 93 0 91 106 102 81
magenta 95 88 91 0 86 91 87
calcium 75 92 106 86 0 49 105
kidney 83 95 102 91 49 0 100
portrait 103 87 81 87 105 100 0
Stupid kidney stones
posted by Alvy Ampersand at 12:57 AM on August 7, 2021
protozoa 0 95 91 95 75 83 103
cupola 95 0 93 88 92 95 87
waltz 91 93 0 91 106 102 81
magenta 95 88 91 0 86 91 87
calcium 75 92 106 86 0 49 105
kidney 83 95 102 91 49 0 100
portrait 103 87 81 87 105 100 0
Stupid kidney stones
posted by Alvy Ampersand at 12:57 AM on August 7, 2021
Your score is 84.57, higher than 83.74% of the people who have completed this task
My words: astronaut, cow, peppercorn, pencil, blanket, seashell, chickpea
Apparently chickpea and peppercorn are closest, followed by cow and blanket. I can see chickpea and peppercorn (small round food items), but why are cows and blankets close?
posted by hurdy gurdy girl at 1:24 AM on August 7, 2021
My words: astronaut, cow, peppercorn, pencil, blanket, seashell, chickpea
Apparently chickpea and peppercorn are closest, followed by cow and blanket. I can see chickpea and peppercorn (small round food items), but why are cows and blankets close?
posted by hurdy gurdy girl at 1:24 AM on August 7, 2021
My fun new game is googling any pair of words that are inexplicably related and learning that, for instance, pineapple teardrop earrings are a big thing.
posted by brook horse at 9:23 PM on August 6
Well this just raises further questions.
posted by ZaphodB at 1:34 AM on August 7, 2021
posted by brook horse at 9:23 PM on August 6
Well this just raises further questions.
posted by ZaphodB at 1:34 AM on August 7, 2021
A friend of mine years ago made up a game we'd play that he called Eyebrows Mushrooms Cough Drops, the goal of which was to rattle out the three least-related things you could think of in the least amount of time possible. In retrospect, I guess cough drops and mushrooms are both things you put in your mouth, but it was a catchy name.
Maybe all that early EMC practice prepared me well for this game:
Your score is 92.08, higher than 98.25% of the people who have completed this task
My least divergent pair (71) were "truncheon" and "windsock". How are *those* the closest two?
The whole list: windsock, truncheon, table, monster, banknote, lubricant, eyebrow
posted by deadbilly at 3:19 AM on August 7, 2021 [2 favorites]
Maybe all that early EMC practice prepared me well for this game:
Your score is 92.08, higher than 98.25% of the people who have completed this task
My least divergent pair (71) were "truncheon" and "windsock". How are *those* the closest two?
The whole list: windsock, truncheon, table, monster, banknote, lubricant, eyebrow
posted by deadbilly at 3:19 AM on August 7, 2021 [2 favorites]
Your score is 89.83, higher than 96.17% of the people who have completed this task
guilt electron hexagon cousin irrelevance abstraction catheter
For some reason it rejected passionfruit, which was one of my first seven, and doesn't seem too technical.
posted by Klipspringer at 3:38 AM on August 7, 2021
guilt electron hexagon cousin irrelevance abstraction catheter
For some reason it rejected passionfruit, which was one of my first seven, and doesn't seem too technical.
posted by Klipspringer at 3:38 AM on August 7, 2021
That's probably because passion fruit is spelled passion fruit.
posted by phunniemee at 3:55 AM on August 7, 2021 [1 favorite]
posted by phunniemee at 3:55 AM on August 7, 2021 [1 favorite]
You're quite right.
*shakes fist at yesterday's Mango, Papaya & Passionfruit yoghurt for misleading me*
posted by Klipspringer at 4:00 AM on August 7, 2021
*shakes fist at yesterday's Mango, Papaya & Passionfruit yoghurt for misleading me*
posted by Klipspringer at 4:00 AM on August 7, 2021
I got 90.21 with banjo, milkweed, fart, curtain, femur, engine, tiara. Strangely, the most similar in this set were “banjo” and “fart”.
posted by transient at 5:23 AM on August 7, 2021 [2 favorites]
posted by transient at 5:23 AM on August 7, 2021 [2 favorites]
Not strange at all.
posted by mono blanco at 5:52 AM on August 7, 2021
posted by mono blanco at 5:52 AM on August 7, 2021
Wish I had known ahead of time that it drops some of the words, using only the first “valid” ones. Mostly because I plugged in “cat” and “toast”, having taken the test while eating breakfast while my cat watched. I mean, I put in “regolith” too, but the test kept “cat” and “toast” instead.
Still scored above 85%…
posted by caution live frogs at 6:02 AM on August 7, 2021
Still scored above 85%…
posted by caution live frogs at 6:02 AM on August 7, 2021
My least divergent pair (71) were "truncheon" and "windsock". How are *those* the closest two?
Fill the windsock with pennies.
posted by solotoro at 6:03 AM on August 7, 2021 [4 favorites]
Fill the windsock with pennies.
posted by solotoro at 6:03 AM on August 7, 2021 [4 favorites]
95.22, higher than 99.5%, using halibut, oncology, epistemology, dwelling, sortie, weekend, cannon. Halibut/weekend and cannon/sortie tied for closest, with weekend/epistemology farthest apart.
I guess we know what people are and aren't doing on the weekend!
posted by mrgoat at 6:09 AM on August 7, 2021
I guess we know what people are and aren't doing on the weekend!
posted by mrgoat at 6:09 AM on August 7, 2021
My least divergent were "photon" and "pinwheel", and now I feel like I have some research to do on why that is.
posted by Zargon X at 7:12 AM on August 7, 2021
posted by Zargon X at 7:12 AM on August 7, 2021
I refuse to believe it - a group as diverse as metafilter readers and they're all scoring way more than 90% of all other people? Either some large herds of facebook readers are taking the test as well, or the numbers are bogus.
posted by DreamerFi at 7:14 AM on August 7, 2021
posted by DreamerFi at 7:14 AM on August 7, 2021
I got 92.72. Most divergent: castaway and commercial, 108. Most similar: dimension and spawn (seemingly because of Minecraft), 80.
I did spend too long thinking of the last three so was sad to see they didn’t count. But I also seemed to get worse as I went on.
posted by little onion at 7:24 AM on August 7, 2021
I did spend too long thinking of the last three so was sad to see they didn’t count. But I also seemed to get worse as I went on.
posted by little onion at 7:24 AM on August 7, 2021
nail, noise, seed, pixel, crater, paper, belt
Your score is 85.8, higher than 87.86% of the people who have completed this task.
Most divergent: seed, crater (95)
Least divergent: pixel, noise (68)
Woot!
posted by Thorzdad at 7:53 AM on August 7, 2021
Your score is 85.8, higher than 87.86% of the people who have completed this task.
Most divergent: seed, crater (95)
Least divergent: pixel, noise (68)
Woot!
posted by Thorzdad at 7:53 AM on August 7, 2021
Strangely, the most similar in this set were “banjo” and “fart”.
Not strange at all.
I remarked to someone just yesterday that when I play banjo I find it enjoyable, but my happiness is rarely shared by those nearby.
posted by ricochet biscuit at 8:30 AM on August 7, 2021
Not strange at all.
I remarked to someone just yesterday that when I play banjo I find it enjoyable, but my happiness is rarely shared by those nearby.
posted by ricochet biscuit at 8:30 AM on August 7, 2021
I happen to have not slept at all last night, and, oddly, I feel certain that it contributed to my high score: ("Your score is 91.88, higher than 98.12% of the people who have completed this task.") I probably thought of all ten of my nouns in about two minutes, which I seriously doubt I would have managed in a more normal state of consciousness, when I tend to overthink things a lot. I wonder if that is evidence that I am in any way more creative when I'm sleep-deprived. It would make some sense; that's the condition in which I wrote all my papers in college, after all.
EDIT: I guess I should post my answers like everyone else is doing, so here they are:
orchestra lemur prison houseboat carton troll puddle
orchestra 0 88 87 96 101 98 95
lemur 88 0 100 98 97 85 87
prison 87 100 0 91 94 86 94
houseboat 96 98 91 0 92 93 85
carton 101 97 94 92 0 99 91
troll 98 85 86 93 99 0 76
puddle 95 87 94 85 91 76 0
posted by davewagon at 8:43 AM on August 7, 2021
EDIT: I guess I should post my answers like everyone else is doing, so here they are:
orchestra lemur prison houseboat carton troll puddle
orchestra 0 88 87 96 101 98 95
lemur 88 0 100 98 97 85 87
prison 87 100 0 91 94 86 94
houseboat 96 98 91 0 92 93 85
carton 101 97 94 92 0 99 91
troll 98 85 86 93 99 0 76
puddle 95 87 94 85 91 76 0
posted by davewagon at 8:43 AM on August 7, 2021
I entered the first ten nouns in the lyrics of REM’s It’s The End of the World and I Know It.
posted by ricochet biscuit at 8:43 AM on August 7, 2021
Your score is 80.34, higher than 63.68% of the people who have completed this taskUnsurprisingly, earthquakes and hurricanes are conceptual neighbours, a mere 43 apart, as are birds and snakes with 41, while aeroplanes and notch don’t even write much these days (107).
posted by ricochet biscuit at 8:43 AM on August 7, 2021
Your score is 92.07, higher than 98.24% of the people who have completed this task
Highest scoring pairs:
Foam — Suffragette (108)
Stud — Suffragette (103)
Foam — Pulsar (102)
Lowest scoring pairs:
Stud — Nostril (75)
Pulsar — Taillight (78)
Nostril — Taillight (78)
Stud—Suffragette has a rather high score, so I assume their corpus has home renovation books but not Harlequin romances.
Nostril—Taillight has a rather low score... are there taillights that resemble nostrils? Are people sniffing taillights?
posted by Monochrome at 8:44 AM on August 7, 2021 [2 favorites]
Highest scoring pairs:
Foam — Suffragette (108)
Stud — Suffragette (103)
Foam — Pulsar (102)
Lowest scoring pairs:
Stud — Nostril (75)
Pulsar — Taillight (78)
Nostril — Taillight (78)
Stud—Suffragette has a rather high score, so I assume their corpus has home renovation books but not Harlequin romances.
Nostril—Taillight has a rather low score... are there taillights that resemble nostrils? Are people sniffing taillights?
posted by Monochrome at 8:44 AM on August 7, 2021 [2 favorites]
I hope they have a secondary project to look at the corpus of words we are all putting in. How many lemurs and catamarans and astronomical objects.
(I got 89.7 on the real go. When it let me go again for funs I got up to 91, but then I stopped trying. )
posted by janell at 9:03 AM on August 7, 2021 [1 favorite]
(I got 89.7 on the real go. When it let me go again for funs I got up to 91, but then I stopped trying. )
posted by janell at 9:03 AM on August 7, 2021 [1 favorite]
Your score is 83.17, higher than 78.05% of the people who have completed this task
baby plaque glitter habituation pie wart jelly
baby 0 87 76 104 69 86 65
plaque 87 0 80 96 88 82 84
glitter 76 80 0 111 76 85 61
habituation 104 96 111 0 100 91 98
pie 69 88 76 100 0 85 50
wart 86 82 85 91 85 0 71
jelly 65 84 61 98 50 71 0
For some reason I was thinking about grad school while doing this. I did research with young infants using a habituation technique. Interestingly, habituation and baby were my second highest at 104. Habituation and glitter was 111... I guess you never habituate to glitter.
posted by kathrynm at 9:19 AM on August 7, 2021 [1 favorite]
baby plaque glitter habituation pie wart jelly
baby 0 87 76 104 69 86 65
plaque 87 0 80 96 88 82 84
glitter 76 80 0 111 76 85 61
habituation 104 96 111 0 100 91 98
pie 69 88 76 100 0 85 50
wart 86 82 85 91 85 0 71
jelly 65 84 61 98 50 71 0
For some reason I was thinking about grad school while doing this. I did research with young infants using a habituation technique. Interestingly, habituation and baby were my second highest at 104. Habituation and glitter was 111... I guess you never habituate to glitter.
posted by kathrynm at 9:19 AM on August 7, 2021 [1 favorite]
I was taught that nouns are person/place/thing/idea, but when I hear "noun" my first thoughts are always "animals". Like dogs and lemurs or whatever. Animals aren't things, so they must be people. (Though certain animals have a special place in my heart, which is both a place and an idea.)
posted by aniola at 9:40 AM on August 7, 2021 [1 favorite]
posted by aniola at 9:40 AM on August 7, 2021 [1 favorite]
I just tried: person place thing idea verb noun adjective
and got Your score is 60.99, higher than 0.54% of the people who have completed this task while also simultaneously being told that my score may be artificially low due to misunderstanding the instructions.
Also their list of gender options is too short and not in alphabetical order.
posted by aniola at 9:43 AM on August 7, 2021 [2 favorites]
and got Your score is 60.99, higher than 0.54% of the people who have completed this task while also simultaneously being told that my score may be artificially low due to misunderstanding the instructions.
Also their list of gender options is too short and not in alphabetical order.
posted by aniola at 9:43 AM on August 7, 2021 [2 favorites]
Your score is 6.2, higher than 0.0% of the people who have completed this task
nine three four five six seven eight nine 0 14 11 7 6 4 5 three 14 0 3 5 7 10 9 four 11 3 0 5 4 8 6 five 7 5 5 0 3 5 5 six 6 7 4 3 0 5 3 seven 4 10 8 5 5 0 5 eight 5 9 6 5 3 5 0posted by reventlov at 10:29 AM on August 7, 2021 [1 favorite]
Eventually I got the score to 98.91 with:
Gelatin
Libertarian
Satellite
Oracle
Purr
Bedsore
Butane
(Along the way, I tried changing gelatin to fruitcake, but fruitcake ended up being too similar to libertarian.)
posted by pickles_have_souls at 10:39 AM on August 7, 2021 [6 favorites]
Gelatin
Libertarian
Satellite
Oracle
Purr
Bedsore
Butane
(Along the way, I tried changing gelatin to fruitcake, but fruitcake ended up being too similar to libertarian.)
posted by pickles_have_souls at 10:39 AM on August 7, 2021 [6 favorites]
You know, everyone *talks* about semantic disassociation, but no one actually purple monkey typewriter. This humor☆☆☆☆☆
posted by Oyéah at 10:39 AM on August 7, 2021
posted by Oyéah at 10:39 AM on August 7, 2021
nice work, pickles_have_souls! I haven’t been able to top 97.78 (99.85%) with bindery lemur disbelief vapor rue wardrobe refinance.
And now I need to stop or I will play this all day.
posted by janell at 10:46 AM on August 7, 2021 [1 favorite]
And now I need to stop or I will play this all day.
posted by janell at 10:46 AM on August 7, 2021 [1 favorite]
Numbers aside, the inverse approach (lowest score) is also a blast. 22.85 (viola, flute, bassoon, piano, cello, clarinet, violin) was the lowest I could get last night.
posted by riverlife at 10:54 AM on August 7, 2021
posted by riverlife at 10:54 AM on August 7, 2021
Your score is 93.51, higher than 98.99% of the people who have completed this task
abacus diatribe manatee interim melamine notary poetess
abacus 0 99 91 104 87 92 89
diatribe 99 0 97 94 100 95 82
manatee 91 97 0 95 96 78 91
interim 104 94 95 0 103 82 108
melamine 87 100 96 103 0 95 103
notary 92 95 78 82 95 0 83
poetess 89 82 91 108 103 83 0
I just realized that the unexpected (to me) overlap between "manatee" and "notary" is that there's a Manatee County in Florida so there are probably lots of documents containing both words.
posted by Daily Alice at 11:01 AM on August 7, 2021
abacus diatribe manatee interim melamine notary poetess
abacus 0 99 91 104 87 92 89
diatribe 99 0 97 94 100 95 82
manatee 91 97 0 95 96 78 91
interim 104 94 95 0 103 82 108
melamine 87 100 96 103 0 95 103
notary 92 95 78 82 95 0 83
poetess 89 82 91 108 103 83 0
I just realized that the unexpected (to me) overlap between "manatee" and "notary" is that there's a Manatee County in Florida so there are probably lots of documents containing both words.
posted by Daily Alice at 11:01 AM on August 7, 2021
All right, after lots of tweaking and optimizing, I have come up with:
Your score is 99.03, higher than 99.92% of the people who have completed this task
phishing 0 95 109 104 100 96 99
exoplanet 95 0 104 92 95 101 96
surgeon 109 104 0 97 94 90 100
novella 104 92 97 0 102 107 103
paraprofessional 100 95 94 102 0 111 98
mesh 96 101 90 107 111 0 88
whey 99 96 100 103 98 88 0
Mesh-whey (88) is obviously the weak link, but I haven't been able to replace one of them with something that doesn't bring down the rest of the score.
posted by brook horse at 11:24 AM on August 7, 2021
Your score is 99.03, higher than 99.92% of the people who have completed this task
phishing 0 95 109 104 100 96 99
exoplanet 95 0 104 92 95 101 96
surgeon 109 104 0 97 94 90 100
novella 104 92 97 0 102 107 103
paraprofessional 100 95 94 102 0 111 98
mesh 96 101 90 107 111 0 88
whey 99 96 100 103 98 88 0
Mesh-whey (88) is obviously the weak link, but I haven't been able to replace one of them with something that doesn't bring down the rest of the score.
posted by brook horse at 11:24 AM on August 7, 2021
This might be my favorite thing ever posted and my favorite thread ever on metafilter. So many people did what I was thinking, or commented what I was considering commenting...
Buttress/butter...108!
My new pup is named butter and I call the pup buttress sometimes and this is apparently a very unlikely association and I am pleased he has such divergent semantically divergent nicknames...
is this secretly a band name generator
It is a band name generator and so, so much more. I mean Mariachi Quark? Glitter Plaque!?
Got "your score is 14.9, higher than 0.0% of the people who have completed this task" by using "one, two, three, four, five, six, seven" then thought "oh wait, numbers are adjectives" and am now down a "what part of speech are numbers" rabbit hole, happy Friday!
I also got this score when I entered the integers. I didn't hesitate over the noun vs adjective status though because I guess a natural number, an integer has to be a thing right? Isn't number the ur-noun almost? What did you discover about the linguistic status of numbers?
Are mefites a uniquely random constituency? I would wager we would outdo most communities in competitive randomness tasks...
This is so beautiful. thank you all...
posted by albion moonlight at 11:31 AM on August 7, 2021 [1 favorite]
Buttress/butter...108!
My new pup is named butter and I call the pup buttress sometimes and this is apparently a very unlikely association and I am pleased he has such divergent semantically divergent nicknames...
is this secretly a band name generator
It is a band name generator and so, so much more. I mean Mariachi Quark? Glitter Plaque!?
Got "your score is 14.9, higher than 0.0% of the people who have completed this task" by using "one, two, three, four, five, six, seven" then thought "oh wait, numbers are adjectives" and am now down a "what part of speech are numbers" rabbit hole, happy Friday!
I also got this score when I entered the integers. I didn't hesitate over the noun vs adjective status though because I guess a natural number, an integer has to be a thing right? Isn't number the ur-noun almost? What did you discover about the linguistic status of numbers?
Are mefites a uniquely random constituency? I would wager we would outdo most communities in competitive randomness tasks...
This is so beautiful. thank you all...
posted by albion moonlight at 11:31 AM on August 7, 2021 [1 favorite]
Your score is 100.2, higher than 99.96% of the people who have completed this task
phishing 0 95 109 104 100 96 105
exoplanet 95 0 104 92 95 101 105
surgeon 109 104 0 97 94 90 104
novella 104 92 97 0 102 107 109
paraprofessional 100 95 94 102 0 111 98
mesh 96 101 90 107 111 0 87
crockery 105 105 104 109 98 87 0
Okay now I'm really done. Really... not going to touch mesh-crockery... not... going to...
posted by brook horse at 11:58 AM on August 7, 2021 [1 favorite]
phishing 0 95 109 104 100 96 105
exoplanet 95 0 104 92 95 101 105
surgeon 109 104 0 97 94 90 104
novella 104 92 97 0 102 107 109
paraprofessional 100 95 94 102 0 111 98
mesh 96 101 90 107 111 0 87
crockery 105 105 104 109 98 87 0
Okay now I'm really done. Really... not going to touch mesh-crockery... not... going to...
posted by brook horse at 11:58 AM on August 7, 2021 [1 favorite]
Oh wow! Pie-Wart-Jelly...not quite thrift store, Cucumber-Sorbet.
posted by Oyéah at 12:23 PM on August 7, 2021
posted by Oyéah at 12:23 PM on August 7, 2021
Your score is 108.11, higher than 100.0% of the people who have completed this task
veterinaries, converts, ruminations, months, swellest, smash, excludes
(well, 108.11359662828701 according to the code I wrote which generated that example). But I needed to get the word list from the author, since the word list in his github repo is much vaster than the one that he is actually using on the site).
Does anyone have an actually-good algorithm for this?
posted by novalis_dt at 1:01 PM on August 7, 2021 [2 favorites]
veterinaries, converts, ruminations, months, swellest, smash, excludes
(well, 108.11359662828701 according to the code I wrote which generated that example). But I needed to get the word list from the author, since the word list in his github repo is much vaster than the one that he is actually using on the site).
Does anyone have an actually-good algorithm for this?
posted by novalis_dt at 1:01 PM on August 7, 2021 [2 favorites]
I am obscurely pleased with my first try:
Your score is 95.65, higher than 99.59% of the people who have completed this task
alpaca borscht law virulence hammer airflow chorus
alpaca 0 90 102 100 90 94 104
borscht 90 0 108 99 103 104 97
law 102 108 0 100 82 100 91
virulence 100 99 100 0 100 84 97
hammer 90 103 82 100 0 89 77
airflow 94 104 100 84 89 0 98
chorus 104 97 91 97 77 98 0
posted by kyrademon at 1:12 PM on August 7, 2021 [1 favorite]
Your score is 95.65, higher than 99.59% of the people who have completed this task
alpaca borscht law virulence hammer airflow chorus
alpaca 0 90 102 100 90 94 104
borscht 90 0 108 99 103 104 97
law 102 108 0 100 82 100 91
virulence 100 99 100 0 100 84 97
hammer 90 103 82 100 0 89 77
airflow 94 104 100 84 89 0 98
chorus 104 97 91 97 77 98 0
posted by kyrademon at 1:12 PM on August 7, 2021 [1 favorite]
"Crock" dropped it to 98.26, unfortunately. It was slightly more associated with every word.
posted by brook horse at 1:17 PM on August 7, 2021
posted by brook horse at 1:17 PM on August 7, 2021
Interesting. When I did it, it gave a 98.68.
phishing 0 95 109 104 100 96 92
exoplanet 95 0 104 92 95 101 97
surgeon 109 104 0 97 94 90 96
novella 104 92 97 0 102 107 100
paraprofessional 100 95 94 102 0 111 105
mesh 96 101 90 107 111 0 86
crock 92 97 96 100 105 86 0
posted by aniola at 1:24 PM on August 7, 2021 [1 favorite]
phishing 0 95 109 104 100 96 92
exoplanet 95 0 104 92 95 101 97
surgeon 109 104 0 97 94 90 96
novella 104 92 97 0 102 107 100
paraprofessional 100 95 94 102 0 111 105
mesh 96 101 90 107 111 0 86
crock 92 97 96 100 105 86 0
posted by aniola at 1:24 PM on August 7, 2021 [1 favorite]
I also read something wrong, I did get 98.68 with that, haha. But yes, still lower than what I got with crockery.
posted by brook horse at 1:31 PM on August 7, 2021
posted by brook horse at 1:31 PM on August 7, 2021
What did you discover about the linguistic status of numbers?
With the caveat that I am not a linguist (and would still love more in-depth analysis of numbers as parts of speech, if anyone has info they want to share on this), I ended up on numerals might be their own part of speech depending on your theory of grammar, and can serve as adjective or noun. So I guess the score holds up!
posted by the primroses were over at 1:37 PM on August 7, 2021 [1 favorite]
With the caveat that I am not a linguist (and would still love more in-depth analysis of numbers as parts of speech, if anyone has info they want to share on this), I ended up on numerals might be their own part of speech depending on your theory of grammar, and can serve as adjective or noun. So I guess the score holds up!
posted by the primroses were over at 1:37 PM on August 7, 2021 [1 favorite]
Your score is 82.95, higher than 77.07% of the people who have completed this task
apple universe patience abstraction field itch crystal
Lowest was apple-crystal and universe-abstraction at 73, highest was itch-crystal at 97
posted by ymgve at 3:09 PM on August 7, 2021
apple universe patience abstraction field itch crystal
Lowest was apple-crystal and universe-abstraction at 73, highest was itch-crystal at 97
posted by ymgve at 3:09 PM on August 7, 2021
Why they don't hire DOCENTs for WIND farms. I saw an article today about passwords, which stated three random words, is better than one complicated word. I can see a little of that type of research here.
posted by Oyéah at 3:25 PM on August 7, 2021
posted by Oyéah at 3:25 PM on August 7, 2021
This was fun!
But as usual with this sort of thing, whether I do well or not I immediately rabbit-hole on methodology and whether the conclusions seem unambiguously justifiable. I did "well" enough for a quick ego validation, but I don't think the specific numbers are all that relevant since just reordering the same set in multiple ways provided a 20 point variance for me.
But I don't think I agree with either the methodology or the conclusions as I understand them from the main text on the website (I didn't read the thesis, so I can't speak to the results in depth, just to how it was represented on the test page and FAQ).
The instructions were to enter ten "words that are as different from each other as possible, in all meanings and uses of the words."
But what they measured was how likely the chosen words were to be used on the same web pages by other people. I immediately suspected that while that might provide some correlation to common contextual use, it completely ignores actual differences in meanings and more importantly, fails to measure the test takers' understanding of those differences. Which is what they had asked for.
So lists of very specific words that are unlikely to be used by the same author at the same time were going to score higher, whether the words are unrelated or not, while less specific, more common words that are actually unrelated in any meaningful way are still more likely to appear together just through statistical usage, and would therefore score lower.
So I rabbit-holed! (Of course!) I experimented.
To test similarity in meanings, I decided to start by grouping words by type. First experiment was types of animals found in subtropical climates. Alike on at least two vectors. I used amoeba, gibbon, hornbill, swallowtail, tapir, coati, mosquitofish, weevil, panther, and shrike. That got a "difference" of 70.74, despite the fact that I only chose animals that the browser's dictionary recognized when I typed them in the form, and many of the animals were listed on the same web pages I used to look them up. (Changing the order didn't significantly change the score for that test.) It wasn't a high score, slightly below average, but I thought it should be much lower.
So I tried another set. Types of tools, but unlikely to be commonly discussed by the same authors at the same time. Key, axe, warhead, sprinkler, dagger, planer, blowtorch, printer, needle, wrench. Score was 83.93, higher than 81.28% of people who have taken the test. I have 7 of those tools in my garage. (Won't say which ones. Never know who's reading this.)
I decided for my final test to use words that are either actual synonyms, or are used in combination. Words that are uncommon enough to not be used in a majority of websites, but have multiple synonyms for me to play with.
I started with one specific word and then played a game of free association until I had a list, then omitted the starting point. (Feel free to see if you can guess what word I used to start this game.) The list was canid, knockout, syncope, propaganda, epistle, dispatch, letter, landlord, hotelier, bankrupt. Result was 88.81, higher than 94.72%, despite that I deliberately chose words that were somehow related to each other.
It's a fun exercise, but I don't think it's measuring what they say it's measuring. It's not measuring the test taker's understanding of the meaning and usage of words, it's measuring the test taker's intuition of everybody else's understanding of the meaning and most likely combined usage of those words. Which is still a very interesting experiment and result. Just... it isn't really measuring you. It's measuring you measuring everybody else.
It's... a meta filter of MeFites filtering semantic dissociation filterers.
MetaFilter: It's... a meta filter of MeFites filtering semantic dissociation filterers.
And like I said: It's fun! Especially since I'm supposed to be working.
posted by It's Raining Florence Henderson at 3:32 PM on August 7, 2021 [4 favorites]
But as usual with this sort of thing, whether I do well or not I immediately rabbit-hole on methodology and whether the conclusions seem unambiguously justifiable. I did "well" enough for a quick ego validation, but I don't think the specific numbers are all that relevant since just reordering the same set in multiple ways provided a 20 point variance for me.
But I don't think I agree with either the methodology or the conclusions as I understand them from the main text on the website (I didn't read the thesis, so I can't speak to the results in depth, just to how it was represented on the test page and FAQ).
The instructions were to enter ten "words that are as different from each other as possible, in all meanings and uses of the words."
But what they measured was how likely the chosen words were to be used on the same web pages by other people. I immediately suspected that while that might provide some correlation to common contextual use, it completely ignores actual differences in meanings and more importantly, fails to measure the test takers' understanding of those differences. Which is what they had asked for.
So lists of very specific words that are unlikely to be used by the same author at the same time were going to score higher, whether the words are unrelated or not, while less specific, more common words that are actually unrelated in any meaningful way are still more likely to appear together just through statistical usage, and would therefore score lower.
So I rabbit-holed! (Of course!) I experimented.
To test similarity in meanings, I decided to start by grouping words by type. First experiment was types of animals found in subtropical climates. Alike on at least two vectors. I used amoeba, gibbon, hornbill, swallowtail, tapir, coati, mosquitofish, weevil, panther, and shrike. That got a "difference" of 70.74, despite the fact that I only chose animals that the browser's dictionary recognized when I typed them in the form, and many of the animals were listed on the same web pages I used to look them up. (Changing the order didn't significantly change the score for that test.) It wasn't a high score, slightly below average, but I thought it should be much lower.
So I tried another set. Types of tools, but unlikely to be commonly discussed by the same authors at the same time. Key, axe, warhead, sprinkler, dagger, planer, blowtorch, printer, needle, wrench. Score was 83.93, higher than 81.28% of people who have taken the test. I have 7 of those tools in my garage. (Won't say which ones. Never know who's reading this.)
I decided for my final test to use words that are either actual synonyms, or are used in combination. Words that are uncommon enough to not be used in a majority of websites, but have multiple synonyms for me to play with.
I started with one specific word and then played a game of free association until I had a list, then omitted the starting point. (Feel free to see if you can guess what word I used to start this game.) The list was canid, knockout, syncope, propaganda, epistle, dispatch, letter, landlord, hotelier, bankrupt. Result was 88.81, higher than 94.72%, despite that I deliberately chose words that were somehow related to each other.
It's a fun exercise, but I don't think it's measuring what they say it's measuring. It's not measuring the test taker's understanding of the meaning and usage of words, it's measuring the test taker's intuition of everybody else's understanding of the meaning and most likely combined usage of those words. Which is still a very interesting experiment and result. Just... it isn't really measuring you. It's measuring you measuring everybody else.
It's... a meta filter of MeFites filtering semantic dissociation filterers.
MetaFilter: It's... a meta filter of MeFites filtering semantic dissociation filterers.
And like I said: It's fun! Especially since I'm supposed to be working.
posted by It's Raining Florence Henderson at 3:32 PM on August 7, 2021 [4 favorites]
This is interesting. I had a party game idea based on word association that's the opposite of this.
Anyways, uh...
"Your score is 86.75, higher than 90.5% of the people who have completed this task"
I guess I think outside the uh... neptune.
posted by symbioid at 3:46 PM on August 7, 2021
Anyways, uh...
"Your score is 86.75, higher than 90.5% of the people who have completed this task"
I guess I think outside the uh... neptune.
posted by symbioid at 3:46 PM on August 7, 2021
Shame you can't do to words, but "fiber optic" I guess would mess up their logic
posted by symbioid at 3:57 PM on August 7, 2021
posted by symbioid at 3:57 PM on August 7, 2021
Your score is 90.54, higher than 96.98% of the people who have completed this task
Aw yeah, first go.
vector plutonium dissolution admiration galaxy quorum fascist
vector 0 94 99 99 81 96 99
plutonium 94 0 75 98 92 99 94
dissolution 99 75 0 86 90 71 81
admiration 99 98 86 0 93 97 73
galaxy 81 92 90 93 0 94 92
quorum 96 99 71 97 94 0 97
fascist 99 94 81 73 92 97 0
posted by ephemerae at 4:06 PM on August 7, 2021
Aw yeah, first go.
vector plutonium dissolution admiration galaxy quorum fascist
vector 0 94 99 99 81 96 99
plutonium 94 0 75 98 92 99 94
dissolution 99 75 0 86 90 71 81
admiration 99 98 86 0 93 97 73
galaxy 81 92 90 93 0 94 92
quorum 96 99 71 97 94 0 97
fascist 99 94 81 73 92 97 0
posted by ephemerae at 4:06 PM on August 7, 2021
if I'd known I was being graded I would have tried harder.
posted by philip-random at 4:39 PM on August 7, 2021
posted by philip-random at 4:39 PM on August 7, 2021
95.28 Calliope/ freezer, 105-xenon/culvert-104, convent/subset 103, convent/xenon 99
Very fun.
posted by Oyéah at 5:13 PM on August 7, 2021
Very fun.
posted by Oyéah at 5:13 PM on August 7, 2021
"program opal stapler fish hour conversation condensation"
"Your score is 91.72, higher than 98.0% of the people who have completed this task"
I feel like I won something.
posted by of strange foe at 6:24 PM on August 7, 2021
"Your score is 91.72, higher than 98.0% of the people who have completed this task"
I feel like I won something.
posted by of strange foe at 6:24 PM on August 7, 2021
On the first try and without reading the thread: Your score is 94.64, higher than 99.36% of the people who have completed this task.
I have a new flex nib pen and I have been writing down words with interesting letter pairs to improve my crappy ligatures, I picked the best 10 nouns from a page and that was it.
Griffin and affirmation scored real well.
posted by Dr. Curare at 11:20 AM on August 8, 2021 [1 favorite]
I have a new flex nib pen and I have been writing down words with interesting letter pairs to improve my crappy ligatures, I picked the best 10 nouns from a page and that was it.
Griffin and affirmation scored real well.
posted by Dr. Curare at 11:20 AM on August 8, 2021 [1 favorite]
I'm not often in the 1% of anything so for my own posteritiness I am putting this down, first try:
Your score is 94.91, higher than 99.43% of the people who have completed this task
piano lemming furlong jaundice concept flag train
piano 0 102 98 106 87 96 77
lemming 102 0 96 93 100 96 90
furlong 98 96 0 103 110 101 94
jaundice 106 93 103 0 103 98 101
concept 87 100 110 103 0 81 79
flag 96 96 101 98 81 0 81
train 77 90 94 101 79 81 0
posted by sylvanshine at 3:44 PM on August 8, 2021 [1 favorite]
Your score is 94.91, higher than 99.43% of the people who have completed this task
piano lemming furlong jaundice concept flag train
piano 0 102 98 106 87 96 77
lemming 102 0 96 93 100 96 90
furlong 98 96 0 103 110 101 94
jaundice 106 93 103 0 103 98 101
concept 87 100 110 103 0 81 79
flag 96 96 101 98 81 0 81
train 77 90 94 101 79 81 0
posted by sylvanshine at 3:44 PM on August 8, 2021 [1 favorite]
ninja warthog cream poet shovel moustache escalator got me an 87 /91% Ninja moustache was the most common link, which I can't see myself, but that's the way it is with ninja I suppose.
posted by Sparx at 7:47 PM on August 8, 2021
posted by Sparx at 7:47 PM on August 8, 2021
Your score is 89.43, higher than 95.64% of the people who have completed this task
electron orange mountain sock roof pawn denouement
electron 0 86 96 103 91 86 98
orange 86 0 72 76 71 88 104
mountain 96 72 0 83 67 94 103
sock 103 76 83 0 87 94 98
roof 91 71 67 87 0 96 97
pawn 86 88 94 94 96 0 88
denouement 98 104 103 98 97 88 0
posted by sebastienbailard at 8:17 PM on August 8, 2021
electron orange mountain sock roof pawn denouement
electron 0 86 96 103 91 86 98
orange 86 0 72 76 71 88 104
mountain 96 72 0 83 67 94 103
sock 103 76 83 0 87 94 98
roof 91 71 67 87 0 96 97
pawn 86 88 94 94 96 0 88
denouement 98 104 103 98 97 88 0
posted by sebastienbailard at 8:17 PM on August 8, 2021
> But as usual with this sort of thing, whether I do well or not I immediately rabbit-hole on methodology and whether the conclusions seem unambiguously justifiable. I did "well" enough for a quick ego validation, but I don't think the specific numbers are all that relevant since just reordering the same set in multiple ways provided a 20 point variance for me.
The program picked the first seven valid words, so reordering would indeed give different numbers.
The program picked the first seven valid words, so reordering would indeed give different numbers.
We use the first seven valid words out of the ten. Using a subset of words allows for some redundancy; you can include up to three invalid words (e.g., misspelled words or words that are too technical) while still receiving an overall score.posted by sebastienbailard at 9:48 PM on August 8, 2021
Apparently Coffee and Nebula get a score of 102 which I call shenanigans on.
posted by Just this guy, y'know at 1:54 AM on August 9, 2021
posted by Just this guy, y'know at 1:54 AM on August 9, 2021
85.85, higher than 88.0%
tree, graph, widget, nation, belief, oration, gulf
enjoyable, will do again.
posted by theora55 at 10:55 AM on August 9, 2021
tree, graph, widget, nation, belief, oration, gulf
enjoyable, will do again.
posted by theora55 at 10:55 AM on August 9, 2021
I had one word combo worth 118. I am fascinated by this. By now I usually hit 96. This is antithetical to how I think. I try to bring things together and even rhyme. But TS Eliot comes to mind, "More distant and more solemn than a fading star/culvert." Calliope / culvert, wind/docent. Cliffie from Cheers voice," Your gonna need a docent for that culvert over there."
posted by Oyéah at 7:24 PM on August 9, 2021
posted by Oyéah at 7:24 PM on August 9, 2021
« Older Sunshine and Ravioli (MACARONI) | Pharmaceutical firms: our for-profit public health... Newer »
This thread has been archived and is closed to new comments
posted by olykate at 4:16 PM on August 6, 2021 [2 favorites]