Postby pir » Wed Aug 12, 2015 10:21 am
Yes, the new male voice in Duolingo's French "course" is better than the female one, but it is still TTS and prone to all the annoying things that come with it. IMO learning good pronunciation from the very start is actually more important than the folks at Duolingo seem to think, so their general use of TTS is a black mark against them. TTS is fine for Google Translate; they are not attempting to teach me a language -- and frankly, GT's French bot is actually better than Duo's.
What I consider even worse though is the weird pedagogy behind Duo's teaching -- the way the "course" is presented to me heavily prioritizes reading and listening. There's no speaking at all (I don't blame them, speech recognition is hard). But there is also very limited production of L2 writing, and that's a deal breaker in my book. I was apparently slotted into the "quick and easy" section of one of their A/B tests, which means that for each unit within a section, I get as few as 7 sentences (which is not enough to learn any new material). Also, 8 out of 10 translations are into my L1. I question the validity of A/B testing for actual learning (as opposed to website design), and this makes the entire endeavour seem like a waste of time to me. I find myself motivated to cheat when "strengthening" sections since I don't actually learn anything much the first time through and when I am asked to produce something moderately complex, I draw blanks so often that it depresses me. Now, I don't cheat because I'm not gonna cut my nose off to spite my face, but I wonder -- if I feel like it, experienced as I am with learning, how many less experienced people will cheat? And that could easily blow their A/B results up without them even knowing it.
Whatever they use as SRS algorithm is complete crap. I often end up repeating sentences ad nauseam that I got perfectly the first time, while sentences in which I made real mistakes (not just typos) don't reappear for ages. Compared to Memrise, this is seriously flawed. Memrise has SRS nailed down; when I finish a "course" there I do know all the words with at least one meaning.
Duo claims I am 59% fluent (I've been gathering that this is meant in regard to the number of words in the tree). That's pretty misleading without any explanation -- I had to search the forums to find some information, and I am still not sure it was correct. They list 1592 words that I supposedly know. I don't. I know (to the point of being able to reproduce them reliably) 552 words. And most of those I refreshed on Memrise after originally learning them decades ago in high school. What basic grammar I know I sought out myself, I certainly didn't get it from the sentences or the short "tips and notes" preceding some of the sections (which have disappeared completely since I passed the halfway mark of the tree). Because of the many errors I've come across in the English translations, I don't even trust the French I am presented with, so are they teaching me good sentence structure, and common ways of expressing myself?
The only reason I am sticking it out -- aside from wanting to finish an entire tree so I can more fairly evaluate it -- is the grammar discussion in some of the comment threads. There are a couple of knowledgeable people who're native French speakers fluent in English, and their explanations beat the pants off my reference grammar. But I've noticed them being less active now than they were a year ago, which probably means they're burning out. That would not be surprising, since the same questions come up over and over and over because the comment threads stand alone (that makes them easy to access from each individual sentence, but means all those gems are buried and most people will never see them).
I see some really nifty stuff happening at Duo, but the basic model leaves a lot to be desired, and I am not sure it can be fixed without throwing a lot of money and experienced teachers at it. Duo proudly touts this study that showed them as more effective than Rosetta Stone (shooting for low-hanging fruit) and a college semester (this was measured by administering a college placement test before and after the study). People did improve. Novices improved the most (predictably), the most experienced people improved the least. But the study says nothing about why and how, and exactly how useful this actually is for people who want to learn a language as opposed to gaining a semester in college. My assessment is that self-motivated self-studiers can do better; I know I could.
PS: "course" in quotes because I consider the terminology misleading. The French offering is not a stand-alone course; it needs to be considerably augmented to actually teach somebody French.
6 x