Typo acceptance/rejection consistency
"Frohes" accidentally spelled "Frohs" -- warned, and accepted.
"Weihnachten" accidentally spelled "Weinachten" -- outright rejected.
Both off by one letter. If anything, the one that was accepted is arguably the worse misspelling of the two (higher percent letters wrong, and even the syllable count is wrong). There should be more consistency (if not forgiveness) in one-letter-off typos.
Thanks for the explaination, but it does not really address the issue. Would you not agree, that Weinachten is clearly a misspelling of Weihnachten? And thus not an attempt to translate christmas to a completely unrelated word.
I realize that it is hard for a computer to judge what is a spelling mistake and what is a genuine misunderstanding, but may I suggest another approach: What if you accept something as a typo if it is of by one letter and it is not a valid inflection of the correct word?
By doing it this way you would still be hard on all the cases where people uses the wrong ending or similar, but I would argue, that in 99% of the cases where people write a word that is of by one letter that is an unrelated valid word, then the error is actually a typo and not an error of understanding.
Take your own example. If someone writes he instead of the the resulting sentence would in most cases be gibberish. What is most likely? That the student thought die meant he or that the student made a typo?
It is impossible for the program to account for all the potential typos. I believe you can use the Report Mistake option and suggest to accept your translation as a typo. (This is only possible if you use the web version).
Overall, if you shift your attitude a little and agree that by default no typos are allowed, you'll be just thankful in cases when they are forgiven :-)
I doubt that it would be accepted if I submitted a typo, and frankly I do not think that it would be desirable from a technical point of view to store all "honest" typos as correct solutions.
I made no mention of my own feelings towards this, and I really have no strong feelings about the subject. My attitude is already "shifted".
But since the Duolingo team already did decide that they want to accept typos that are off by one letter I am merely suggesting another way of doing it, and arguing that doing it the way would more closely approximate what a human would consider an "off by one letter"-typo and what they would consider an error of understanding, thus making it more consistent from a human point of view.
Last I heard, every sentence accepted has to be input by hand. This can generate hundreds if not thousands of alternatives.per sentences. Limitations, such as those listed by Luis, are necessary to reduce the otherwise intimidatingly extensive possibilities for typos for each sentence.
No, the "off by one letter" errors that Luis describes are most likely decided automatically by the program and this would also be possible to do with my suggestion.
Thanks for the info. That sounds like the right way to go, however I cannot find a translation for "weinachten".
In German compound words are written in one word. So most likely the program thought that you meant to combine the word wein and achten. That is at least my best guess...
(two years later) As a German native speaker I'd say the second typo is worse! Weihnachten without the h is quite a common spelling mistake as there is no difference in the sound of the word but how the word is derived. Missing the e in Frohes is no doubt just a typo (if not on purpose as some kind of slang).
Interesting point I hadn't thought of. So my typo (wein and achten together) could mean the eighth wine, I suppose.
That would acutally mean "to pay attenttion to wine". From the context everyone will for sure know what you mean, but loads of Germans are quite eager about correct spelling ;) (the eigth wine = der achte Wein)