|
 |
How many English words do you know? | ||
Share with:
| ||
 Language Learning Forum : Specific Languages |  | |
 Iversen Super Polyglot Moderator Denmark berejst.dk Joined 6703 days ago 9078 posts - 16473 votes  Speaks: Danish*, French, English, German, Italian, Spanish, Portuguese, Dutch, Swedish, Esperanto, Romanian, Catalan Studies: Afrikaans, Greek, Norwegian, Russian, Serbian, Icelandic, Latin, Irish, Lowland Scots, Indonesian, Polish, Croatian Personal Language Map | Message 9 of 21 30 January 2014 at 10:25am | IP Logged | |||||
|
It may for all sorts of mathematical reasons be problematic to take your positive score on real words and subtract the number (or percentage) of false positives on non-existing words. But it is also a problem because it assumes that you know precisely X words, and that you always can separate known from unknown words. But that's not how things are in the real world.
Ideally you should add the number (or precentages) of words you understood perfectly, but mistrusted because you couldn't say with certainty that you had seen them. I skipped several correct answers fearing the the punishment for false positives, but didn't get any bonus for doing so. In my own latest statistics I have a category for 'guessable words', and because I take the words from a dictionary it is fairly certain that they all exist. But sometimes it is difficult to see the difference between 'known' and 'guessable'. One afterthought: at the polyglot conference in Budapest in 2013 Anthony Lauder (aka Splog) claimed that one of the main differences between polyglots and polynots is that the former are very adept at guessing. Isn't it then a bit problematic to punish guessing harder than not answering a question? I'll have to look at the d' index to see whether it really "is independent of your response bias - a number that doesn't get affected by whether you guess a lot or not at all", as stated by Patrickwilken. Edited by Iversen on 30 January 2014 at 5:19pm 1 person has voted this message useful | ||||||
| daegga Tetraglot Senior Member Austria lang-8.com/553301 Joined 4521 days ago 1076 posts - 1792 votes Speaks: German*, EnglishC2, Swedish, Norwegian Studies: Danish, French, Finnish, Icelandic | Message 10 of 21 30 January 2014 at 10:55am | IP Logged | |||||
Now what does that mean exactly? :) I also responded yes to a few words I didn't know but which seemed very likely to be an English word. There were a few of those I wasn't sure enough, so I pressed "no". If the question was purely about discriminating words and non-words, I'd have pressed "yes" more often. edit: 2 more runs as a discrimination test: 70/10 as a vocabulary test (have I seen this word before?): 39/0 The 39/0 seems a lot more realistic to me than the 60/0. Edited by daegga on 30 January 2014 at 11:28am 1 person has voted this message useful | ||||||
| patrickwilken Senior Member Germany radiant-flux.net Joined 4533 days ago 1546 posts - 3200 votes Studies: German | Message 11 of 21 30 January 2014 at 11:42am | IP Logged | |||||
That's exactly why SDT was developed. The simple intuitive idea that you either know something with either 100% or 0% is just not plausible. SDT assumes informationally noisy internal representation.
Check out any article on "Signal Detection Theory" and you should get a reasonable explanation of d' and other measures. The whole rationale of SDT is to create sensitivity measures independent of response bias. What should a sensitivity measure do? It should indicate how well you can say a signal is present when it is (Hit Rate), and additionally how well you can say a signal isn't there when it isn't (1 - False Alarm rate). Hit rates on their own are useless, because you don't know how much of the Hit Rate is composed from lucky guesses. So you measure the guessing rate independently (the False Alarm rate) to allow you to adjust and create a 'pure' sensitivity measure. Guessing is not punished. The whole idea is that decisions are essentially noisy and you can't avoid making some false positives and some false negatives when making a judgment. It's assumed that your sensitivity is fixed (here your ability to distinguish words from non-words). The only thing you can change is your bias/criteria/threshold when you say one or the other. Imagine: (1) I'll give you 100 Euros for every word you correctly identify in the test; or (2) You'll give me 100 Euros for every non-word you incorrectly classify as a word. In 1 you'll make a lot of false alarms (i.e., saying non-words are words) because you are trying to maximize your winnings (ideally you'd say everything is a word). In 2 you'll be very careful not to make mistakes and so your False Alarm rate (saying a word is there when it isn't) will be very low. In both cases your underlying knowledge of words (your 'sensitivity') stays the same, just your criterion bias changes. SDT was developed to try to try to develop a pure measure of sensitivity. It's not perfect, but it's a lot better than a simple percentage correct. --------- The main point for me is simply that I know from experience in running visual experiments, that d-primes that are larger than 2 are close to ceiling. Basically if your d' is that high tests are not useful at distinguishing performance (same is true if d' is very close to zero). A d' of +2 implies you have no difficulty distinguishing the two stimuli (here 'words' and 'non-words). The another thought is that this test is really about estimating your ability to distinguish between two stimuli classes ('words' and 'non-words') - not directly about how many words you know in English. Your sensitivity will obviously depend on the nature of the non-words. If all the non-words appeared in red, and all the words in blue, then presumably everyone would be able to get close to perfect performance. I don't know how the 'non-words' were generated, but presumably there are certain assumptions being made here about what a realistic non-English-word would look like (random letters would presumably be way too easy). It could be that the experiment is assessing the sorts of internal models of English words that native and non-native speakers have, by systematically manipulating certain structural properties of the non-words (e.g., perhaps non-native speakers are more likely than natives to think that words that end in "logy" are real words for instance). It might be for that people score highly not because they know all the 'words', but that their internal model of what an English word is is really good (or at least really good at distinguishing between the non-words on the test). Edited by patrickwilken on 30 January 2014 at 2:17pm 2 persons have voted this message useful | ||||||
| dampingwire Bilingual Triglot Senior Member United Kingdom Joined 4665 days ago 1185 posts - 1513 votes Speaks: English*, Italian*, French Studies: Japanese | Message 12 of 21 30 January 2014 at 2:13pm | IP Logged | |||||
I've just checked in the OED. "instable" the adjective, meaning unstable, is marked as rare. "instable", the verb meaning to stable a horse, is marked as both rare and obsolete. Interesting that they have such words in their list. 1 person has voted this message useful | ||||||
| beano Diglot Senior Member United KingdomRegistered users can see my Skype Name Joined 4622 days ago 1049 posts - 2152 votes Speaks: English*, German Studies: Russian, Serbian, Hungarian | Message 13 of 21 30 January 2014 at 2:40pm | IP Logged | |||||
|
Practicable is another word that exists in dictionaries but is rarely heard in real life. (I've heard it once, and assumed the guy had made a mistake).
1 person has voted this message useful | ||||||
| Ogrim Heptaglot Senior Member France Joined 4639 days ago 991 posts - 1896 votes Speaks: Norwegian*, English, Spanish, French, Romansh, German, Italian Studies: Russian, Catalan, Latin, Greek, Romanian | Message 14 of 21 30 January 2014 at 3:15pm | IP Logged | |||||
I think you point to one of the weaknesses of the test. I got a score of 76/0, but obviously I said yes to several words for which I would not be able to give a clear definition or a translation into another language, I just "knew" by intuition that it had to be a real word. On the other hand, this does mean that if you see those words in a context, you are likely to get the meaning by how they are used. So as a word-recognition test it has its merits, but obviously it doesn't mean that you have an active vocabulary of 50.000 words in English, the way I read it. 1 person has voted this message useful | ||||||
| daegga Tetraglot Senior Member Austria lang-8.com/553301 Joined 4521 days ago 1076 posts - 1792 votes Speaks: German*, EnglishC2, Swedish, Norwegian Studies: Danish, French, Finnish, Icelandic | Message 15 of 21 30 January 2014 at 5:36pm | IP Logged | |||||
Was this guy per chance a German (or French) native speaker? "praktikabel" is not uncommon here, and I would use it in English without a second thought to be honest. Good to know that I shouldn't. Edited by daegga on 30 January 2014 at 5:43pm 1 person has voted this message useful | ||||||
| Medulin Tetraglot Senior Member Croatia Joined 4668 days ago 1199 posts - 2192 votes Speaks: Croatian*, English, Spanish, Portuguese Studies: Norwegian, Hindi, Nepali | Message 16 of 21 30 January 2014 at 6:52pm | IP Logged | |||||
|
''You said yes to 71% of the existing words.
You said yes to 7% of the nonwords. This gives you a corrected score of 71% - 7% = 65%. This is fairly high level for a native speaker. Also help our colleagues who are investigating word associations. Nonwords you responded YES to Seconds ; ; slitly 11.999 &n bsp; tertiell 8.539'' Edited by Medulin on 30 January 2014 at 6:57pm 1 person has voted this message useful |
|  | 
Printable version
|
| Copyright 2024 FX Micheloud - All rights reserved No part of this website may be copied by any means without my written authorization. |
