1. Forum
  2. >
  3. Topic: Chinese
  4. >
  5. Are there still devs working …


Are there still devs working on Chinese?

I've seen people write in the discussions over three months ago about answers that ought to be correct that still haven't been added. Are people still working on fixing the Chinese course or have they abandoned it?

It gets pretty frustrating after a while when you have to remember what Duolingo wants you to answer rather than what is correct.

March 18, 2018



The only changes I have ever seen made on the Chinese course after launch, are alternate translations being accepted. I have reported extensive bugs and other reports, but I have yet to see any of them fixed or acknowledged.

These include:

  • Wrong or missing hover-over hints. This is a problem throughout the whole course, from the very beginning, and is sometimes severe, making it very hard to get through lessons without using the word bank.
  • Characters with missing or wrong pronunciations, which can lead to some strange things in matching exercises. There aren't many of them but the fact that they have gone months without being fixed bothers me.
  • Strange or unnatural behaviors with the algorithm, like getting the same sentence repeatedly, or having it give you a practice session that consists exclusively of matching characters to phonetics without any translation.
  • Deep pedagogical defects in the course like very long, complex sentences appearing earlier on, forcing you (like you reference in your comment) to memorize the "correct" answer just to get through the lesson...and then later sections having very simple sentences...rather than an organic increase in complexity of sentences like many of the older courses used to have (at least before some of the recent updates like how the German course got royally screwed up).

I would expect the course to be incrementally tweaked and improved over time, but I see no sign that it is being worked on other than the alternate translations being accepted. (And this is being carried out by volunteers, not by developers or paid staff.)

This is one of the many reasons I cancelled my Duolingo subscription. I care about the Chinese course and I want to see it reach the level of the other courses, but IMHO it is nowhere close, and I see no sign that any progress towards improving it is being made, or even that any of the Duolingo staff cares about it at all and are putting any effort into it whatsoever. There is not even any communication about it...zero.


For the wrongly-pronounced characters are they wholly wrong, or just wrong for the context in which they are taught? I don't think the Japanese course has figured out how to resolve the analogous issue, and it's been out threeish times as long. I think it probably requires a deep coding fix.

The repeated sentences seem to be a problem in every Duolingo tree. I haven't seen the other problems in your first or third bullet points personally (and haven't gotten far enough to comment on the fourth one), and certainly not in comparison to what passes for hints in many other courses long out of beta.


I think it probably requires a deep coding fix.

It clearly requires a coding fix. But if this is a "deep" fix, and not something that is fixable quickly and easily, I'd question either the competence of the current developers, or the way the system is designed and set up to begin with.

I'm a web developer and something like this would be trivially easy to implement...just add a new field in the database to whatever table in the database is being pulled from to generate the characters for those exercises, and have that field specify an "override" for prononciation, and feed that override into the TTS and not the character itself, using the character itself in the case that the field is empty.

It doesn't take a genius to come up with the solution. This fits in the category of "Things I look at and immediately know how to solve without having to think about it."

In the extremely implausible scenario that it's not possible to program an override like this for some bizarre reason that I can't imagine...a slightly more labor-intensive solution is to program a manual override and have it play a separate audio file which can either be provided by recording the TTS in a controlled setting, or if that for some reason fails, having a human actor that sounds vaguely like the TTS record it.

How much does Duolingo pay their software engineers? Surely they can hire people smart enough to think up and implement such fixes.

The fact that it hasn't happened, to me, suggests that it's just not being prioritized. It's sad if you ask me, I can't imagine a fix like this would take more than minutes to program, less than an hour to test and roll out. I have rolled out much more complex bug fixes in various projects that I've worked on, in a much tighter time-frame than this.

Having it go on months and months like this without being fixed (and without any other bugs being fixed), makes the product look like abandonware. The whole Chinese course looks like dump-and-run, to me, from a software development standpoint. Zero follow-up.

You point me to a SINGLE software-level bug fix or programming change, post-launch, for anything related to the Chinese course and I'll happily backtrack on my words, but I have yet to see ANY examples.


I'd question ... the way the system is designed and set up to begin with.

The simple fact it took dozens of programmers months to do the necessary work to make the Japanese (and, by extension, Chinese) course possible, I think, yeah, this, and I bet the people working on it every day are also like, yeah, this.


For the wrongly-pronounced characters are they wholly wrong, or just wrong for the context in which they are taught?

I've never come across a wholly-wrong reading, but I have seen a few quite obscure (and thus not very useful) ones; they are always technically-valid readings, though, and presumably arise from the TTS not being designed to present the most common reading of an individual character devoid of any context. As DL is not responsible for designing the speech engine, I imagine the only fix would be to manually re-assign all the erroneous sound files.


Answers are still being accepted. I have received two emails in the last month. Just be patient and report.

In the meantime, just try your best to remind the preferred answers. And for other issues, utilize resources like Google and other websites to help solve any other learning problems that you may have.


The more likely scenario is that I quit using Duolingo. Haha!


I hadn't gotten any for a long time, but now I got 25 in the past month. There's at least a few working on it.

I've seen noticeable improvement in some parts of the course, but there's still a handful of lessons that are a lot worse than the others translation-wise.


One area that could be cleaned up right away is to accept equivalents for 'very' in the English translations. E.g. 'really' 'truly' 'extremely'. Also, placement of adverbial phrases of time, e.g., today, tomorrow, now in English is very flexible, but gets rejected too often by DL.


Perhaps you'd like the "reverse course": English for Chinese speakers. It'll be a bigger challenge and has also had years longer to get missing translations added. Of course, most people in that course aren't reporting missing English translations.


Yes, there are. I have gotten six emails from Duolingo reporting that they've accepted suggested sentences.


I've received almost 30.


I got a Chinese suggestion email only a day or so ago, so they are certainly still implementing alternative translations. In general, I feel that the course is very much improved from its initial condition in this respect, even though there are still gaps.


The thing is that for us that haven't been here since the beginning we don't have anything to compare with. A study tool that marks a correct answer incorrect isn't a good way to gain new users. It would've been one thing had it still been in Beta, but it's not. I can't help but feel like it has been rushed.


Unfortunately crowd-sourcing alternative translations takes time, and the Chinese course simply hasn't been around for long enough yet. I do agree that DL graduates courses (in general) from Beta too soon, however, although this might be, in part, an attempt to encourage more users to a course which will, in turn, expedite its improvement. 'Beta' is an arbitrary designation, anyway, as all DL courses differ and some are of much better quality than others.

To give an idea of the improvement, in the Chinese alpha test, it frequently only accepted a single English permutation for a Chinese sentence. There were some skills where I was submitting reports on literally every question. ‘你們’ was universally translated as 'you guys' and rejected 'you'...
Yes, the course still has plenty of faults, but carry on submitting reports and I'm sure they will eventually be ironed out.


I decided to no longer use the Chinese course. Not only do you get trapped in modules by not remembering DL's awkward English translations and having your English translations improperly rejected, but I find that the teaching approach is based on rote memorization of characters and sounds completely divorced from meaning. New vocabulary is presented as a character and a sound, without any reference to what it means or how it is used in a sentence. After a couple of months of daily use, I feel that it wasn't adding anything to my daily Chinese language study - teacher led course, online course material, and Anki flashcards.


I feel the same. The English the machine understands is a little limited


I've had many suggested sentences accepted in the last few weeks.


My problem is that the difficulty does not seem to increase with increasing crown levels.

I levelled Greeting1 up to 4 because I wanted to test myself on listening, recall and pronunciation, but the most difficult I ever got was "translate to english" and "match character with sound". In fact the only difference between the crown levels 2 to 4, and all of their subsections was the length.

Spanish is a lot better in that regard.


It seems to me that the Chinese course simply doesn't have some of the exercise types that one might expect to find at higher skill levels (notably writing down of a sentence read in Chinese). Undoubtedly, this is having an effect. Given how limited Basics 1 is in Chinese, I don't think it can probably give a comprehensive picture.

The Chinese course is new and seems to have a relatively small inventory of sentences, so I wouldn't be surprised if the effect of skill levels is less visible. Hopefully more sentences will be added as time goes on.

Learn Chinese in just 5 minutes a day. For free.