posted by murrayjames on May 30, 2012 | 8 comments
I've noticed that the spacing on user-generated pinyin questions is quite strict. I got these answers wrong:电话 dian4 hua4十块 shi2kuai4不懂 bu4dong3My pinyin was right, but not the required spacing: dian4hua4, shi2 kuai4, bu4 dong3. I imagine spacing wouldn't matter, since in Chinese word boundaries are less important than getting the spelling and tones correct.
I've noticed this as well. The English definitions can also be a little tough for things like 顺便 where it needs to be 'in passing, conveniently' instead of just in passing. It'd probably be pretty long winded to fix things like this as that is how the definitions are programmed into the backend database that likely serves up content. Unsure what the best fix is, but it could use some tweaking.
Yes, being marked wrong, as I was, for answering "have confidence in" instead of "TO have confidence in" comes across as a tad pedantic, though is likely just a database thing... Can you incorporate some kind of Natural Language Processing function?
@craigrut and @half_wasted, OK... have just put in some quick changes to help with these issues: (1) the spacing issues with pinyin input are fixed, (2) for definitions split by a comma or semi-colon, providing either of the listed definitions is sufficient, (3) for arbitrarily long words, we'll now mark your answer as correct if what you provide is a long enough exact match against a portion of the correct answer. Hopefully these will make the generative exercises a bit less frustrating by avoiding marking answers as incorrect when people really know them and it is just an input issue.
Cool, I'll give them a shot later today. Thanks for the quick switch!
It still insists on "TO" before verbs, else says answer is wrong.
We've just loosened the requirements here. The system will now also be a bit more relaxed in handling singular and plural nouns, so "second" or "seconds" will both work for 秒, etc.
Another small wiggle with pinyin input... I'm studying both 得 dei3 and de2. I normally don't mind that no definition is provided for generative pinyin questions. Unfortunately, with 得 (or any 多音字) it's impossible to know which answer is correct absent context or definition. EDIT: Is it possible for pinyin questions to have more than one acceptable answer? I have a similar problem with 啊. Though the lesson says 啊 a4, I keep answering 啊 a1 -- valid pinyin, incorrect answer. EDIT2: Ditto for words where the last character can be neutral or not, depending.
A few other pinyin points.Every so often an answer will have the first syllable in neutral tone, eg. mei5miao4, xiong5mao1. It's easy to discard these wrong answers, since 5th tone almost always appears at the end of a word, not the beginning.Slightly more common is impossible pinyin combinations, like pun, zuai, etc. I'm not sure how complicated the algorithm for matching pinyin initials and finals is, but it could use some fine-tuning.