Ask HN: Feedback on Skritter, Learning Chinese with Spaced Repetition on App Engine
skritter.com
skritter.com
When we were first getting started, App Engine was just becoming available, and we thought that would be a great platform mainly because of its scalable architecture and Python backend, which both Nick and Scott liked. We incorporated the Spaced Repetition after we read about Supermemo in Wired Magazine shortly after launching.
The idea is to charge users small monthly subscriptions with big discounts for semester-signups. Since a lot of our current users are young and in school it's really important that we keep the price affordable. We've also attended 2 trade conferences in language learning software, and will be going to one more in about a month at which we'll be trying to sell Skritter as a site license to institutions.
If you'd like to check it out in more detail, we've set up a demo account for everyone on HN:
UN: hackernews pass: hackernews
Thanks a lot for your time.
I remember being poor and in college once, where "poor" means I was never in danger of starvation but I did routinely decline invitations to the movies for want of $5.
I also remember a collection of someones (parents, university, scholarship providers, government, self) feeling it was worth $80 per hour of instruction to get me Japanese classes. Please do not use the "Oh noes, our customers are poor!" factoid as a mental excuse to undercharge for your product. ($80. An. Hour.)
I'd also like to note that a significant portion of the American collegiate population has a significant amount of discretionary spending capability. Half of my Japanese class had parents who were paying full freight for that $80 an hour. Do you really need to sell to the other half?
So how much does it make sense to charge, from an outside perspective?
Your customers are not buying online SRS, they are buying an opportunity to improve their Chinese proficiency, and the experience of being the kind of person who wants to improve their Chinese proficiency. Never forget that.
I have a friend with an as-yet-unreleased application for studying Japanese. I told him what I'll tell you: I wouldn't consider charging less than $30 but I would consider trying more than $100. Yep, that is quite a range, and it all depends on what type of student you are targeting. To be perhaps a little unfair to my friends' prospective customers, I would expect to see no anime-related vocabulary in the $100+ product.
Exactly how you present your choice to the user is up to you. ($49 a semester, $80 if you buy a year at once should lead to an ARPU in excess of $100, hopefully.)
We actually do have some users who put together a kung cu list for us... uh oh!
Luckily, I don't think there is any place for ad-financed products (ads are too distracting when you are trying to memorize stuff like this), so there won't be many free commercial systems.
I am building something very similar to this. I have to say I am very impressed with what Skritter has done so far.
Are you acquainted with Rosetta Stone? They sell language learning software in the "couple of hundred dollars" range. And they're pretty freaking big.
it is difficult to differentiate
People always tell me that about my product. It isn't true for me, and my product is substantially more boring and smaller in scope than learning the key to accessing the history/culture/economic potential of China. A short list of points to differentiate on would be who you sell to, what level of learners you target, what focus of learners you target, visual design, etc etc.
There are definitely many interesting aspects of this market. I don't think it is impossible to differentiate or I wouldn't be in it. But, I really do think that prices for formal educational books, materials, and software are all heading towards $0 fast.
Rosetta Stone really has executed brilliantly in its segment of the market. It sales and marketing effort has been top-notch. It is also interesting to see how they are attempting to cross over from do-it-yourself language learners to traditional language learners. I would love to know what they are charging schools for their software.
We are making our own set of language data, which is a lot of work but which does help differentiate because students can see we have their textbook.
I always wonder how similar "very similar" is. Of course I'm curious. We haven't seen anything that (we think) is similar yet.
edit: p.s. our idea was rejected by yc, so maybe it is a terrible idea. we still think it is great which is why we are working on it.
http://news.ycombinator.com/item?id=548299
My co-founder noted that the rejection email contradicted this, but I'm sure there is some there is an explanation such as we are bad at selecting successful startups. It is tough to really know for sure, because while some of the groups selected fail, some of the ones who ought to have been selected will never succeed on their own when they may have with some guidance. So don't blame yourself, there are two parties involved and neither are God, just human.
ps/ I am interested in learning Chinese, however I am not sure I would pay at that point... Need to see incentives/success stories.
Good luck guys!
Yeah, we really do need an instructional demo video on the "try it page." We actually just got the front page up, so new instructional demo videos and the like are on the way. It's good to know people actually want them, and sorry for any confusion.
Regarding you interest in seeing success stories or incentives, we are going to conduct a study next semester with a big batch of students so that we can make claims about the efficacy, is that the sort of thing you're looking for ("With Skritter, you can learn character 44% faster than with traditional methods, see our study") or would it be more enticing to see small narratives from users about how much its helped them?
What are your criteria are for verifying the validity of the study? I've proofread lots of graduate papers about studies of second language acquisition and in almost every case the studies are summarized the same way: the results are (un)promising but no definite conclusions can drawn because the sample size was too small and/or there was too strong of a selection bias and/or there were some other significant flaws in the study. A study that would justify such a bold claim would be, as far as I can tell, a landmark study in the field--likely worthy of a Ph.D. People could probably get Ph.D.s just for studying your study! :)
I think many people will be interested in the results of your study. Do you know when/where the results will be published?
The Supermemo technique should be good at keeping the pace up to get to that magic 6000 character level. Once you get a lot more data I would think you would have something worth publishing in journals of education and that could be a very nice PR opportunity for you, showing that your software dramatically improves the time needed to become fluent in Chinese.
6000 characters is very extreme. I think 3000 is a more magic target.
I expect we can produce some very sexy stats, once we build a data analysis framework for it. Do you think we could publish just based on our online practice data, without doing a more traditional study?
As for publication, I'm really not sure. I think it would depend on you finding someone else's data for Chinese language learning for comparison purposes. There's probably some conference somewhere you can present at, if nothing else.
I've been using your app for months. This is a fantastic way to practice character writing.
Simply phenomenal work.
Thank you!
[update: I figured out how to use Select: List. Strangely it's taking a long time to add the list. The HSK1 list should already exist on your server somewhere, no? Maybe it's doing it character by character? Is this an App Engine-ism? I would think a simple UPDATE student_character_list INSERT '?' WHERE source_list='HSK1'; would be pretty fast (if I even got that SQL right) but then maybe App Engine doesn't support SQL?]
I just created an account. Is there a difference between the functionality for my fresh account versus the hackernews account, other than different usage data?
Yeah, App Engine doesn't do SQL. Makes things interesting.
No difference in functionality--everyone is the same right now.
http://en.wikipedia.org/wiki/Chinese_schools, http://www.ncacls.org/ncacls_frm_intro.htm
These community based organizations often rely on eastern teaching methods (often rote) to instruct large numbers of young children. Your app would definitely introduce a different mindset (and fun!), when learning Chinese.
Btw, will you guys support different dialects (ie. cantonese)?
Right at the moment we don't support any other dialects due mostly to development resource constraints, but Cantonese is on the radar. One thing I didn't mention in my opening comment was that we are going to expand it to Japanese in the near future, which should widen the number of people interested.
Good: Everything that happens inside that beautiful white square. Stroke recognition, hinting, etc. The demo video is also good.
Regarding the UI clutteri-ness, are you referring to the flash interface or the webpage surrounding it?
UI, I'm referring to the webpage around the flash. It's a little hard for me to tell because I'm on FF2/Linux now and the site is kind of janky, these are some things I'd change:
- On the homepage, the buttons don't look like buttons.
- On the homepage, the way that the black line follows the borders of the buttons is ugly. Maybe it's the aliasing. If you're going to have that line at all, the horizontal should align with the vertical center of the buttons, instead of some arbitrary point.
- On the learning interface, the horizontal lines extending to the left of the flash are totally arbitrary looking.
- Too many thick black lines. Use contrast to control where the user's attention is drawn. Right now the front page goes logo -> description -> buttons/nav-menu (ignoring the video). Should be description -> buttons -> nav-menu -> logo. Part of this should include making the buttons and nav menu look more clickable.
In a fluent writer, each stroke follow from muscle memory, relative to the size and position of the previous stroke. Repositioning breaks the rhythm.
I don't know, the positive reinforcement might work very well for a beginner. I would be more worried about the intermediate student and above.
We're thinking of not repositioning strokes, or maybe not even recognizing them, until each component is completed. Will improve recognition heuristics, give less hints, and be a natural way to indicate components. We have t build the decomposition system first, though. Might that solve the problem you describe?
You might try recognizing the strokes, but leaving them in the background in a lighter color?
[big edit here]
Maybe resize and reposition the model character around the input character as it is written, and then provide some sort of visual feedback at the end on the delta? eg, good proportions, but you are 25% too small and 30% off center.
I'm trying to think of some way to a) preserve feedback while b) reinforcing the rhythmic quality of writing. Rhythm is a mnemonic aid. If you can find some way to reinforce it, that would be a huge boon to students.
There's a rhythm to each stroke and a composite rhythm to each character. Combined with regular patterns like radicals and phonetic roots, rhythm is what allows you to memorize thousands of characters. To the point that sometimes you have to write a character to remember its stroke order.
[/edit]
Very cool site, though. I studied all through the 90s, and the pedagogy was just awful. If I knew then what I know now about language acquisition, I could have compressed my studies into 2 or 3 years, rather than 10. You should have an open field.
Also: I'd love to see a system like this for 行书. The heuristics would probably be a lot more complicated, though, and the market much smaller.
The video is excellent... I just sat there and stared at it. Only criticism is, being able to draw the characters on-screen does not automatically translate to being able to write them by hand. Do you have a way to translate the on-screen writing skill to manual writing skill? If not, I would suggest this simple method: provide a way for the user to select characters that he/she has already learned on-screen and print out a practice sheet (A4 or Letter sized, probably as a PDF), filled with measured squares (with the guidelines), with the first square containing the character, the next 10 containing a pale grey version of the character, and all the other squares blank.
In my experience I needed to fill most of an A4 page before being able to write the character freehand without guidelines - perhaps even more than that. If you have this feature, then next time I'm learning chinese (!) I'll gladly pay an "affordable monthly subscription" (like, about $5-10) to use your tool, probably for about a year (I'd probably buy a year in advance).
Oh, as someone pointed out, you'll lose all the students with that subscription... perhaps worth offering a student subscription for a little less (or even contacting universities to subsidise their student memberships?).
Good luck!
You're right that writing in the screen only isn't going to do anyone's handwriting any prettiness, but so far it looks like knowing how to write each character transfers from one to the other. Perhaps calligraphic style can be learned by interested students on their own?
Just as a bit of background, I am fluent in Chinese, have a degree in Chinese, and a few years ago, published a set of Chinese flashcards that are being sold in China, Taiwan, Hong Kong, Australia, Japan, the US, and soon, Europe. I have a lot of relevant experience and knowledge to contribute.
However now I'm not a absolute beginner any more so I have a reasonably good intuition about stroke order, but if I write them in the wrong order I guess I don't really care very much.There are already many good spaced repetition programs like Anki, which are free. I've implemented my own system based of Leitner flashcards for the Nintendo DS (I haven't released it), which is what I use.
Skritter's deal is to make spaced repetition easy (premade textbook lists, no need to grade yourself) and efficient (taking advantage of domain-specific knowledge), while building writing practice into it. Certainly if you're making your own SRS, you probably don't need the ease of use. Your Nintendo DS app sounds cool--are you planning on releasing it, or is it a personal project?
How do you deal with radicals and learning related characters?
We don't have a way of linking related characters together. Haven't thought of any good ideas for it, yet.
As an aside, how did you accomplish those great perspective transformations in your front-page screencast?
To answer your question about the demo video, I used Adobe After Effects and 3D layer manipulations of video captured in Growler Guncam to create the video effects. We used Guncam, which is a pretty out of date capture program mostly intended for capturing in game footage, because it has the highest capture fps unlike camtasia and its ilk. The problem with Guncam is that it's really resource intensive at 24fps, so we were really restricted in how much of the screen we could capture. And just for total disclosure, I figured a lot of this out myself, but the web developer for fontstruct, who I contacted when building our demo video, was very helpful in guiding me to the right methods and I owe him one. He definitely deserves a shout out.
Edit They are currently asleep. I will let them know in the morning however. Thanks for the tip off.
The only website I know that could lookup Chinese characters based on strokes drawing on screen is http://www.nciku.com/
About GAE & GFW, I strongly suggest skritter to support httpS protocol fully
I'd like to learn a simple character as soon as I've signed up.
Also, it may be sweet to add little quiz games to Skritter. Like timed strokes, or something fun (you appear to be creative) to really entice people and motivate them to move on to the next lesson. It always feels better to conquer something like that in an online learning app... And finally, please tell you guys are planning an iPhone app?
An iPhone app will be awesome, but we're pretty bottlenecked on developing power, so it'll have to be a while. Too bad!