Decode Chinese: We are building the future of language education (Kickstarter coming soon)

I’m 90% sure your post is bait :wink:

My post? No, it’s not. Why, what do you mean?

OK. Long answer here, not as the mod, also not aimed at individuals. but rather ideas and what was said. I am direct in communication (Autistic) and I do have some experience in the Chinese learning and teaching area.

How is the actual role of the instructor changed in this product? An AI teacher is not significantly different from a human teacher in terms of what happens to the student’s brain unless its method is different. If you’re saying “this replaces humans with a convenient online thing”, okay. But that really doesn’t supercede anything out there at present as humans are all over teaching online for pay and for free.

Also, what “university research project”? A senior undergraduate thesis? A paper for a class that grew in concert with some techy friends? Or a project with participation from actual professors who are creating research designs, etc. etc. and looking at trifles like sample size and what can actually be said based on the data? I get not publishing, but usually one would proudly state the university affiliation and tout one’s qualifications to create that kind of revolutionary product. I think people are reacting to that omission.

You have significantly more faith in the quality of user-provided contributions than I do. You may want to take a glance at what has happened to the “KudoZ” system on a translators’ site called The idea is that a user submits a question, ‘professionals’ (aka ‘members’) answer it, and then the user picks the best answer which stays on the site as a resource for reference. Half of the answers are crap. It (the ProZ thing) is a great idea from the perspective of promotion and (the significant number of) site users who are naive to the nuances of the language, but as a tool for professionals looking for accurate language, not so much.

Yep, like schools do now. Having “ownership” is more a matter of motivation and/or satisfaction, not a guarantee of quality. Who is determining what is quality content? With a teacher, (theoretically at least, lol) that person knows the language. You’re counting on people who don’t know the language to provide language. That’s one of the huge issues with traditional teaching methodologies – the potential for bad input or student-to-student “throughput” that is the non-proficient teaching the non-proficient.

Progression of what? Memorization of words? I can’t conceive of a way that this system would actually be evaluating acquisition of syntax in a automated yest meaningful way. Yes, it can put the words the student missed more than four times in red font through CSS, but memorization isn’t how the majority of people become proficient in a language. So far all I’m hearing is “we’re getting content from the users, and our system will track their results on memorizing stuff” which is pretty much what’s out there, other than the idea that naive-to-Chinese users are providing your content.

Okay, great! But what does this personalized learning tool provide that is superior to a flash card system? If you come on a forum like this and claim you’re building the future of language education, people want to know.

I’m sorry, but I think the best idea would have been to curb the enthusiasm until at least the promotional materials are ready. So far I’m hearing a lot of buzzword and claims in the text, but basically the underpinning is “a database can teach languages” and “the database content will come from random people”. Greatly simplified, but perhaps not too far from the truth. I think (speaking as someone who worked on a small team on a project to use data from multiple languages to flow into “the same system”) you may have some expectations that will remain unfulfilled when you get to the point where you’re hoping that languages that are hugely different typologically will behave similarly.

I’d also like to know what methodologies of language instruction you are providing through this system. And what the credentials or backgrounds of the designers are in the language field, not coding or tech. (Though of course those are also important.)

I think with the Forumosa crowd, who take this kind of thing pretty seriously, you might have been better to just say “Here’s a new alpha website we’re working on, see what you think.”


“Catering to” individual pace of acquisition and learning styles in a class situation is precisely what teachers are trained to do (okay, probably not the ones that work part-time in a typical Taiwanese buxiban teaching Mandarin; I’m talking about certified teachers). Certified teachers in an actual school are continually evaluated and differentiation is a huge part of that. The differentiation is way more effective when the teacher adopts a mastery-based method, which can sometimes be blocked by administration for “reasons”, but it doesn’t mean AI is the only thing capable of providing that.

1 Like

Thank you for your post. Please give me some time to reply to all of your feedback and questions. We are currently getting a designer onboard to improve our marketing materials and help to clarify our message on the website and for further announcements.

In retrospect, this thread was perhaps a bit impulsive in our attempt to get feedback and exposure early on. However, at the same time, it is important to come to that conclusion by bringing our project from our closed user tests to the general public and evaluate the situation. Ultimately, we are primarily engineers and linguists, not marketers and salespeople. Hope that is understandable and please be patient. I noticed at least some of the questions you had have been answered in some of my previous posts above such as regarding our university affiliation and funding.

// Alexander, Decode


I’m not saying it can’t be done to an extent, but can a teacher give 15 hours of dedicated, perfectly crafted instruction to an individual in a classroom environment? Impossible. Only rich people who got private teachers can do that. AI can give that to everyone.

Sure, take your time. I’m primarily a linguist and language teacher/learner, not a marketer, as well. So I just want to know the facts, not the marketing. I’m really interested in whether this is really going to be something different or not. I’m a little concerned in my own mind at this point that this is a project from linguists, not teachers. There’s a big difference, and a lot of input from teachers, even if the point is to replace them, would probably make it more effective as a self-teaching product, if that’s what is envisioned.

All I’m seeing about university affiliation is a statement that there is one, and a list of the financial backers. I don’t see a university name nor a “San Zhang, Ph.D, lead linguistic architect” or anything like that in this thread. I may have missed something, wouldn’t be the first time, but that’s what I came away with.

Anyway, thanks for the response. It’s a point in your group’s favor that the dialogue is continuing here when people raise questions. I think we’d like a bit more specificity (you’d be surprised what linguistic and engineering things people here do understand) but it’s nice to see someone not just posting once and never appearing again. :slight_smile:


The difference comes in what kind of instruction is given. 15 hours of dedicated, “perfectly crafted” instruction means what, exactly? What’s the language teaching background you’re looking at this from (not buxiban, I mean)? That might help me understand why you think AI is perfectly crafted in and of itself. I’m happy to let you see what my learners can do after X hours of (imperfectly crafted) group instruction (though since I teach online at the moment, it’s not 30 learners at once as is typical in public schools, since I start with a limit of 8).

I think you maybe have a somewhat overidealized view of how AI would be superior to ordinary teachers who have training, other than perhaps availability or cost. AI doesn’t eat, only the company owners who run it do. :wink:

This appears to be a reference resource (dictionary and a growing database) combined with something similar to an advanced flashcard system. I’m sure there’s some interesting stuff in it and it might possibly be useful as a supplement alongside other learning methods. Unfortunately, however, I have to concur with @ironlady : I don’t see how a student using this as his or her sole or primary learning resource could efficiently acquire proficiency in understanding native level speech and writing. The student is being exposed to and words and example sentences, but as far as I can see, is not being exposed to extensive comprehensible input in audio and text in the target language.

I did my postdoc in the field of education technology, and I was disappointed to see so much “old wine in new bottles”: the same old inefficient techniques being put into apps, V.R etc and being backed up with spurious, albeit peer-reviewed, research. I would love see a single app or website for Mandarin learners (or for Taiwanese learners of English) based on comprehensible input, integrated with internal dictionary/translation/explanations, starting at the absolute beginner level and bringing the learner up to a level where they would have acquired a vocabulary of many thousands of words and mastered the main structures of the language: the learner would then be in a position to start seeking out and enjoying material aimed for native speakers. Creating such an app/website, while demanding care in the preparation of engaging, difficulty-appropriate texts and accompanying audio, would not demand any cutting-edge technology or AI systems. There are already excellent programs of this sort for some languages, but as far as I’m aware there’s nothing ideal for Mandarin Chinese. A free program of this sort would also be a great contribution towards education equality, and the cost of producing such a program quite small: students of all ages could learn a language by themselves, and they’d be learning more efficiently than if they enrolled in a language course at any of the world’s leading universities.

I believe that a properly thought-out program/app for self-learners, based on comprehensible input, could be superior even to the best classroom learning, and certainly incomparably superior to your average classroom language-learning experience. An independent learner, using a suitable program, is continually self-adjusting his or her pace, sensing whether he or she needs to move on to new material or review, in a way that is impossible in a multi-student classroom. With the boon of smartphones and ear-cancelling headphones, an independent learner can also find more time to expose themselves to the target language: five minutes here and there on the subway, at the gym, walking the dog, etc.

By eschewing classroom learning in favor of independent learning based primarily on comprehensible input, people like me have able to relatively painlessly acquire genuine proficiency in not just one but many foreign languages, including languages often thought to be difficult for English speakers like Russian, Japanese Chinese. It’s just unfortunate that many so novice learners, even after deciding on self-learning rather than expensive, relatively-inefficient classroom tuition, are waylaid into choosing ineffective and gimmicky apps by means of which they will never acquire fluency in a foreign language.


Hello @jinyu and thank you for your feedback.

I must say that your reply is the first in this thread that elegantly captures most of the core issues and not just the surface symptoms of modern language education (offline and online). I think you will be quite satisfied to see how close your ideas are aligned to our upcoming presentation, design, and marketing materials.

However, you might also be surprised to realize some deeper problems we have discovered after conducting tests with hundreds of students and multiple teachers at three different universities in Taiwan alone. If convenient, I would like to personally present and discuss our research to you before we officially publish. Just send me a DM and we can get connected.

// Alexander, Decode


Hello ironlady and thank you for your questions and feedback. I will provide a better explanation when our new presentation is completed.

Regarding our academic background, the team have changed since the beginning but I think we have a section on that at the end on our website. But in summary:

  • Alexander ( M.S Industrial Engineering & Computer Science with international Chinese language path, Linkoping University & Fudan University)
  • Jonathan ( B.A Computer Science & Linguistics, Harvard University & CS/Information Engineering, National Taiwan University )
  • Jessie ( M.A Applied Linguistics, University College London & Teaching Chinese as a Second Language, Ming Chuan University )

Regarding our university affiliation, we are funded and supported as a research innovation project by Linkoping University (Sweden). The university innovation project started before our current company. It was not until about two years ago that we after the proof of concept had been validated that we incorporated with continued support and funding from the university.

Hope that clarified some of your questions. As I said earlier, we will announce a better presentation, design, and materials created by a marketing specialist which will be more suitable than the confusing website and video I created myself as an engineer.

// Alexander, Decode

1 Like

Can I keep my ears, please? :disappointed_relieved:


If one wants to enjoy the ambient sounds of a walk in the park or whatever, great! Far from me to want to take away someone’s zen time. But if you’re learning a language, you’re going to need countless hours of audio input in some form. And using noise canceling headphones while on the metro or wherever, as long as the outside decibel levels are not too extreme, allows you to squeeze in more hours of input into your busy life.

Noise canceling is fine. I just don’t want my ears canceled! :sweat_smile:


I’d like, if I may, to make a couple of general points about possible pitfalls of research into language learning methods.

The first is that the research might show improvements in a particular learning metric, but without showing the degree to which improvement in that metric parlays into real world linguistic competence. Scientifically rigorous research might show, for example, that a particular flashcard system allows students to memorize word lists faster than another flashcard system or faster than traditional route learning methods like writing words out by hand. But such research, in itself, would not show the degree to which memorizing word lists (for instance) improves a student’s ability to understand spoken and written Mandarin, which, along with speaking, is what the student cares about. There would be evidence for the effectiveness of the Decode program if it were shown that students had, using only the Decode program, (leaving aside speaking for the moment) efficiently acquired the ability to comprehend previously unfamiliar authentic Chinese audio and texts.

The second is the question of the benchmark for competition. I think most of us here are all in agreement that a lot of traditional classroom teaching methods are terribly ineffective. Research might show that students using only Decode for, say, 200 hours, indeed acquired a higher level of genuine comprehension than students who spent 200 hours in a classroom where students are doing things like memorizing characters by writing them out by hand. But that’s not the benchmark for competition. @ironlady has proposed one alternative: classroom instruction by an effective teacher based on comprehensible input and not old-school route learning and analysis of grammar. I proposed another, the method I myself have successfully used in acquiring various foreign languages: independent learning based on massive comprehensible input. In the case of both of those contenders, most of every hour spent learning the student is being exposed to natural Chinese speech and/or writing, not to words or sentences in isolation with accompanying explanations in English. Again, I can see that Decode might be very useful as a reference tool (looking up details of Taiwanese usage, looking up details about a character etc etc) and useful as an effective flashcard system (for people who - I personally don’t - find it beneficial to supplement their learning with a little bit of flashcard memorization). What I don’t see is how, lacking as it does complete and progressively structured texts and audio, Decode could possibly compete as the sole or primary learning tool with the two contenders I have just mentioned. I’m happy for you to PM if there’s anything you wanted to discuss - language learning is an interesting topic and it’s good to see new new pedagogical approaches being discussed. Nevertheless, again, I don’t see how Decode could possibly compete with these two contenders based on the criteria that really matters: not performance on particular metrics, but overall, real-world comprehension and speech.

I realize now that I’ve ignored the question of learning characters which is often cited as a particular difficulty in learning Chinese. And to be fair to the Decode program, I believe it claims to reduce the difficulty students face in learning to read characters? That’s an issue element of Chinese language learning that might need to be discussed separately (my own hunch is that the best way to master reading thousands of characters is possibly simply to go back over dialogue/stories one is already familiar with as audio and pinyin, reading them now with characters. But that’s an interesting an discussion in its own right - can the phonetic elements in Chinese characters be leveraged for faster learning? Is there a role for mnemonic methods? Etc. Etc.)

Good luck with that living in cacophonous Taipei :wink: I once met a severely hearing-impaired guy. He came to Taipei and liked the city. His one complaint: all the noise!

100% agree. I also think immersion is the best way to learn a language. Let the tv run with Chinese in the background, listen to Chinese while doing sports and so on. Live in Taiwan/China and go out to speak with people. The hardest is getting to B2 level - from there it becomes easier (but will still need loads of time). The goal is not how can I minimize the amount of time I spend on learning Chinese - but how can I minimize the time I would consider as work.

For me the best approaches are chinesepod to listen while commuting/doing sports - while watching chinese drama for more active learning.

But yeah - I still haven’t found a way to learn characters that feels fun and not like work. I cannot sit down and rote learn - that’s something I really hate and cannot motivate myself to. I thought I could learn characters by watching drama and at some point just learning it from the chinese subtitles - but 3.5 years into learning Chinese as above - I feel I still haven’t arrived there - and I will need to kickstart it somehow else by learning 2000-3000 characters differently.

Yes I know my method is not efficient. I have once met a russian guy, that had gotten to HSK6 levels from 0 within 1 year - and writing his PHD thesis at Beijing Daxue in Chinese. Crazy. He did it by 16 hours full focus on Chinese per day. But for most people that is simply not feasible. I have a company/job to run besides learning Chinese - so I cannot commit to such a thing. Learning Chinese must happen in my “free” time. So mainly while doing sports or commuting. However I am lucky to be flexible to work from everywhere - so moving to Taiwan/Chinese helps a lot.

My biggest hurdle is that Chinese tv content is really not attractive to me. And then I cannot read enough to find interesting vloggers from China/Taiwan easily in topics that interest me. Likely I should have long ago put in some months of learning characters - it’s just that I hate them so I always postpone it to next week…

1 Like

I find it very interesting that you feel that @jinyu is addressing the “core issues” and I am not, since we are saying the same thing.

Your app shows no evidence of doing anything instructional other than drilling words. There is no provision of comprehensible input, which is the mechanism by which language is acquired, other than bilingual reference entries. That’s how you learn a language for which no other resources are available, and usually not to proficiency or fluency.

But what do I know, I only have a PhD in teaching Chinese, MAs in Chinese linguistics and conference interpretation, and 20+ years of teaching and teacher training. No, I probably wouldn’t have anything useful to say about the “core issues” in your research before it’s published.


For Chinese literacy, “cold character reading” may be of interest.
((PDF) Cold Character Reading: A Chinese Literacy Strategy | Diane E Neubauer -

We’ve been working on this approach, starting with beginners, since 2011 (I’m “Waltz”, as cited in the journal article). Probably not directly useful for advancing students in the 1500+ range, and we assume no required handwriting of characters from memory (since I cannot find anyone who is still required to do that extensively or without references available in today’s society, other than students), but we’re producing emergent readers whose reading behaviors and results are quite different from what has traditionally been produced using memorization.

We are still interested to know where the “sweet spot” in adding memorization lies with respect to writing characters from memory, not in regard to reading. The point at which I would add memorized characters to the mix for reading would only occur after students had acquired the majority of the structure of the language, which in CI terms in a group class could be around 100-150 hours (I’m estimating but that seems close.) At that point, they’re not going to lose the thread of meaning in the previous part of the sentence because decoding and comprehension of that part will be automated, so there’s sufficient working memory to stand up to a quick “lookup” into memorized (or possibly glossary) symbols.

Repeated encounters with characters in context is always going to be more valuable than repeated encounters with characters in isolation, at levels before all the major syntax has been acquired, because of the additional exposure to syntax, collocation, and so on. So extensive reading at or below level is going to be more globally valuable than any flashcards.

1 Like

Why not eavesdrop? Great thing about living in Taiwan, the world is my classroom. True, some of the input is less comprehensible, but difficult to beat it for immediacy and practical application!