Will I be able to do fluent IPA transcriptions of english by the end of 2023?
10
456
190
resolved Aug 24
Resolved
YES

Currently learning IPA with an ANKI deck and want to become more fluent at transcription.

Resolution criteria:
If, at any point before 2024, I manage to:

1. pick a random youtube video with american english in it (sorry, british phonemic IPA is dogwater)
2. choose 10 random seconds of the video containing a reasonable amount of speech

3. transcribe the speech in phonemic IPA (in contrast to phonetic transcription (which might have aspiration and distinguish between other free variations))
4. compare my transcription to IPA pronunciations sourced from wiktionary

5. use `1 - (Hamming Distance) / (Length of string)` to determine a percentage score

6. score is better than 90%

then this market resolves to YES.

Otherwise this market resolves to NO at close.

Will attempt to record my evaluation attempts with OBS for proof.

Get Ṁ1,000 play money

🏅 Top traders

#NameTotal profit
1Ṁ422
2Ṁ20
3Ṁ4
4Ṁ3
5Ṁ3
Sort by:
bought Ṁ500 of YES

random vid selected: https://youtu.be/UJeSWbR6W04?t=1037 10 seconds from here


my transcription:

az mʌtʃ soʊlər æz meɪ, wiər jʌst lərnɪn wɛðər pætərns aʊt hɪr θru græfs ʌv səʊlər dʒɪnəreɪʃən ə ðɪs ɪz ɑlsoʊ əraʊnd ðə seɪm mʌnθ ðæt aɪ gɑt ði eɪt slip soʊ ði eɪt slip ɪz eɪ pɑd kʌvər ðæt fɪts ɑn ə mætrɪs


subtitles:

as much solar as may we're just learnin weather patterns out here through graphs of solar generation uh this is also around the same month that i got the eight sleep so the eight sleep is a pod cover that fits on a mattress


wiktionary sourced pronunciation:

æz mʌtʃ soʊlər æz meɪ wɪər dʒʌst lərnɪn wɛðər pætərns æʊt hɪr θru ɡræfs ʌv səʊlər dʒɪnəɹeɪʃən ə ðɪs ɪz ɑlsoʊ əraʊnd ðə seɪm mʌnθ ðæt aɪ ɡɑt ði eɪt slip soʊ ði eɪt slip ɪz eɪ pɑd kʌvər ðæt fɪts ɑn ə mætrɪs


i didn't type the affricate ligatures on purpose (not on my keyboard)

ər is equivalent to ɚ just a graph difference

ɹ and r are a matter of preference

learnin' pronunciation not in wiktionary so i altered lərnɪŋ to lərnɪn

ʌ and ə are exactly the same sound but stupid RP rules have bled into GA IPA to dictate which one is when

there was one i->ɪ that i got really mad about and almost edited the wiktionary entry but i forget which one

i have the caught-cot merger don't yell at me


205 total character length

az->æz (first letter i wasn't warmed up)

jʌst-> dʒʌst (still wasn't warmed up)
and one other legitimate error but i forget where

3 mistakes out of 205 so i think we're good, feel free to yell at me if you disagree with my methods and we can argue about phonetics.

i kept a screen recording of the process if anybody doesn't trust me but I had to get up in the middle of it to go bring my partner something, so it's like 30 minutes long.

predicted YES

@brubsby suffice it to say i now think i have more than adequately fulfilled the spirit of the market, which was basically "i want to know how IPA pronunciation works for english", and I confidently feel I do now, so I'll be resolving YES. Thank you everyone!

predicted NO

@brubsby Congrats!

predicted YES

@brubsby kudos for the excellent followup, and most of all for going through with your goal!

predicted YES

drafted up some resolution criteria to make things a bit more concrete here. please let me know if any market participants take issue with the criteria. 10 seconds may seem like it's a bit short. but the average "phonemes per second" in english is around 10-12. so we can expect the samples to be around 100-120 characters in length, which seems plenty.

predicted YES

@brubsby since I'm holding YES shares to motivate myself, I wanted to make sure the criteria was moderately stringent (for fairness) while not being too arduous to administer, although looking everything up on wiktionary will probably be the most annoying part. (maybe i'll use gpt-4 to write a program that does this part for me).

predicted NO

@brubsby Hmm… I think I would have expected “fluent” to be a higher bar than 90%.

predicted YES

@JimHays is 95% agreeable? I imagine putting it too close to 100% is just going to result in me having to repeat the test a bunch due to basically meaningless intricacies of phonemic transcriptions, whereas the spirit of the market was to get me to be able to produce correct pronunciations when reading english IPA on wiktionary.

predicted NO

@brubsby I’m not familiar enough with the domain to know how much variation to expect in a task like that. You’re free to set the bar where you think is reasonable.

Using school grading as an anchor I think of 90 as being the bar for an A-, while I would expect fluent to be an A or A+, but that might not translate well to the described task. As a NO holder I won’t turn you down if you’re offering to raise the bar to 95!

Thanks for coming up with a specific definition!

bought Ṁ20 of NO

everybody is too confident in my ability to do this, diminishing my desire and incentive to actually learn it :'(

bought Ṁ200 of NO

@brubsby you've got no chance of doing it, betting the market down. You can prove me wrong

bought Ṁ200 of YES

@firstuserhere (with incentives now aligned) YOU'RE WRONG I CAN

bought Ṁ100 of NO

@brubsby Hahaha we'll see!

bought Ṁ10 of YES

Currently workin' on a QMK keyboard layer with some english phonemes, I wish keybr or something had a good typing training with phonemic transcriptions though.

predicted YES

@brubsby figured out my QMK keyboard layer for phonemic transcription, works pretty good! been trying it out by adding IPA transcriptions to wiktionary, here's a related goals market for that effort:

Does fluent mean in real time, or just without looking up phonemes of your target languages?

@Duncn certainly not in real time, but I think "being able to transcribe arbitrary English without looking up phonemes" is the resolution criteria.