Yet Another Knowledge Estimator

[imron]

Definitely should limit the length of the test. I really got bored in the end.

But obviously that will affect the accuracy. I imagine it's a fine line trying to balance them out.

[Artem]

I definitely agree that it will affect the accuracy. However, I was on 125/135 with no end in sight. It just went up every time 121/131 122/132 etc.

At least, it would be nice to know a decent estimate of how much is left. If it say 125/175, I would probably finish, but not knowing how long to go, I just didn't want to waste time.

[bozhidao]

Artem: Thanks for the feedback. I'll see if I can reduce the number of questions necessary to produce an estimate for very knowledgeable users in the next version. I don't know nearly that many words myself, so it's helpful to hear about your experience.

querido: No need to apologize! Thanks for trying it out

imron: Bingo ;)

Bo

[Artem]

I think it's okay to have a longer test, but maybe have a better estimate of how far you have to go would be helpful to keep going.

I just got a feeling that the test wouldn't stop unless I started missing words.

[chinopinyin]

@bo

It's really great having the frequency lists you used. I would even link them in the about section of your website.

You can call them the Bo Baike Baidu (BBB) frequency lists of the Chinese Language. I am sure many people will be interested in them. These are broad-based frequency lists of 21st century written Chinese

[animal world]

A few observations. Most people here use a SRS program so they probably know quite accurately how many characters and words they know. I also suspect anyone studying Chinese with any degree of commitment will eventually stumble upon a SRS program. If it weren't for that, i would have been thrilled to know where i actually stood. As is, i took the test to validate the accuracy of your program. For the character test it's quite accurate.

I had the distinct sensation that this test was set up to "punish" me the more i knew (it reminded me of the anti-intellectualism of Bush and Palin). It started out with 20 questions and the more correct answers i provided, the bigger the number of questions became. "Yep, you got another correct answer! Now you have more questions to answer."

It may not be easy to do but i was wondering whether instead of starting with zero you could start with say 2000. If people don't know, you give them easier questions and more difficult ones for those who give a correct answer until the program can accurately estimate the score. Imho, this character test will be particularly useful for beginners knowing 500 characters or so. It might give them an impetus to continue studying the language. It might also be very useful for people who didn't keep up with Chinese for a while and want an assessment of what knowledge they retained before diving back into studying the language.

I tried also the three character and four character tests and stopped after 4 in each test. By then, i had figured out that i could probably beat the test all day long by the process of elimination. In fact, i could easily imagine some of us here doing a silly contest about who can get the highest score in the 4 character test

Another refinement that would be welcome is to make some of the choices trickier and include choices that will make people trip up more easily. For instance, for the word 领导人 the 4 choices given were: guǎn lǐ xué, Jī dū jiào, lǐng dǎo rén and lǎo nián rén. 人 is a very easy character that you learn in week1 and right of the bet you can eliminate two choices. Of the 2 choices remaining, lǎo and nián are also easy words. So, you're left with lǐng dǎo rén and will select that even if you might not know 领 and 导.

[Artem]

Also, the four character test appears to have a lot of two character adjectives with two character nouns. I don't think those are actually 4 character words.

[creamyhorror]

I gave up at 200 two-character words, all of which I either knew or guessed correctly. Went by elimination a few times.

I think it might be better if an option to self-judge was available: i.e. the word is shown and you're allowed to select "i know it" or "i don't". It's too easy to figure out the answer given the choices, and it's probably a fool's errand making the choices more similar.

Good job on the system, though, it's a nice tool!

[chinopinyin]

Another Knowledge Estimator

http://www.clavisinica.com/character-test-applet.html

It may be interesting to compare both

My impression is that estimates are not very accurate for people with small vocabularies. Unfortunately, it is possibly more important to know whether you know 400 or 700 characters that to know whether you know 14400 or 14700.

[abcdefg]

Another Knowledge Estimator

http://www.clavisinica.com/character-test-applet.html

It may be interesting to compare both

I just tried both and Bo's was "kinder" by which I mean easier to guess when uncertain of the answer. Clavis is "trickier" and I get the feeling it's trying to trip me up.

With Bo's, I kept wondering when I would be through and finally just quit after 200 two-character words. Bo's has lots of useful words that Clavis doesn't.

[creamyhorror]

The Clavis one was trickier for me because it's single-word and I did two-word for Bo's. It also seemed to pick similar pinyin choices quite often.

First try: Got eight of 72 wrong, for an estimated total of 2945 characters. But I guessed 6 correctly despite not really knowing them, so it should be lower. Second: eight wrong again, for a total of 2867 characters, guessed 3 or 4. I guess the actual number of characters I know is probably 2500 or less.

[skillphiliac]

-

[Sarevok]

My scores

1) 6039 / 3514 / 3725 / 3366 / 3653

2) 8800 / 9976 / 10866 / 15816 / 19391

3) 6297

4) 4197

Clavis

3442 / 3514

When I was first taking the single character test, I correctly guessed quite a few characters based on their radical and phonetic components, hence the score of 6039, which is a bit off the mark. When taking the test again after that, I didn't attempt to guess and just clicked submit, when I didn't know the characters. I estimate my characters knowledge to be somewhere between 3500-4000, so all those other test results (and those two from Clavis) were in the right ballpark...

The 2 syllable compound words test is not nearly as accurate, as there are way too many proper nouns (names of cities and people I have never heard of), which van be correctly guessed without much effort, leading to unrealistic scores like 19391 When I passed them, the scores seemed a bit more realistic (8800-10866). I estimate the total size of my vocabulary (altogether, not distinguishing between the number of syllables) to be around 13000 or so - well, that's the total number of items in my SRS and it doesn't mean I know every single one of those perfectly (my retention rate is at around 87%). But there are some words I know, which are not included in there (those basic ones, I would say the whole HSK 甲 level and a portion of 乙), so I think it could even out at that number...

Tests for 3 and 4 syllable words are completely off (not only proper nouns here, but also quite a word combinations, which shouldn't be classified as 4 syllable words in my opinion)

[valikor]

The test said I know 837 Hanzi (probably about right--not guessing)

Just for fun, I'm doing the 3-character test and guessing on all the questions. I think I might be able to go on for several hours by using process of elimination. I wouldn't be surprised if some of the people on this forum could actually go on for days if they carefully guessed on each question -_-

EDIT: Finished the test.. I apparently know 6,000 3 character words. I should have done the 4-character test, which would probably be easier? Good fun.

Edited January 31, 2010 at 01:10 AM by valikor

[chinopinyin]

What a pity http://bozid.com/quiz seems to be down

[Ed Log]

http://www.clavisinica.com/character-test-applet.html

I think I know approximately 20 characters but just scored over 200.