Комментарии 13
Amazing analysis and serious intentions.
I wish you not to repeat fate of FluentForever app of being payed beta (perhaps, forever)
I wish you not to repeat fate of FluentForever app of being payed beta (perhaps, forever)
Personally, I've been thinking about starting my own learning app for quite a while, basically aiming to create «better Anki» with decent usability.
And I have some considerations and ideas that you missed or just might find interesting. They are based on personal experience of using several language learning apps and also on analysis of issues with Anki (reported and discussed on reddit and support forum).
====
1. There are a lot of people learning subjects other than languages, notably — medicine. They also need a good app with basically same requirements (SRS). Targeting this audience would extremely broaden community of users.
They, however, have some unusual demands regarding cards format and interaction (ordering items, image occlusion, matching pieces, etc).
My own concept is to move cards presentation/interaction completely to web-layer with API to core scheduler, and support creating cross-platform addons implementing any custom cards using web technologies.
2. It is crucial to make decks easily shareable. In spite of principle «your own cards work better» there exist usecases when people prefer to use prepared decks (notably «starter packs», or book courses converted), to collaborate on decks, or to provide new cards from teacher to students.
My idea is to separate card content from learning statistics, and to keep content as simple zip archives of yaml-based files along with all assets. This would allow to share decks just like normal files (without stats) and in non-obscure format, and also to use some CVS systems for advanced collaboration.
3. There is unimaginable variety of ways users want to import and modify their cards and collections. Most noticeable meta-usecase is importing from various other apps (like memrise and quizlet), csv files, emacs orgmode, etc. Occasionally some people need to do some batch processing like global search/replace or reorganize tags.
My idea is to use yaml-based format as a base for storing cards content. This would allow to use good old text editors to edit cards manually, and also to implement format-specific features in advanced editors like Vim, Emacs, Atom.
====
Also, regarding learning mechanics.
1. Quite obviously, any interaction works better than just recalling, because it involves substantial activity, and makes stronger neural response. The «tapping test» is wrong because it simplifies and actually reduces that activity (from typing to clicking).
However, the format of «word ordering» could be useful if used properly. Maybe some cases of pure grammar tasks (like Spanish direct/indirect pronouns, or German modals/infinitives and separable verbs), or in pure ordering tasks (like sorting aminoacids by polarization).
2. Many things are better to learn based in contrasting. Well known are minimal pairs for phonetic. The same is applicable to grammar, in form of choosing correct form. This way requires two interactive features: making choices and preview how answer «looks like» in context.
Selecting predefined choices is also considered a bad practice, but it still makes sense when task focus is to find some matching structures, and all possibilities might be not known to learner.
3. There's no need of negative response to errors like big red message with disgusting sound. It's quite possible to learn without any «stick» part.
4. Images intensify imagination and help to build more bright usage context. Even if they are not strictly related to definition. But they can easily be annoying when selected by other people. That idea of built-in image search by Gabe Wyner seems quite awesome, if implemented properly.
====
Not long ago, I have implemented Interactive cards for Anki, that demonstrate some of the features I wish to have in an app.
The cards allow interactions: typing, editing, reordering words, selecting options. User answers are usually visible in context before submitting. And there's no negative response — in case of wrong answer they show correct version and highlight it to attract attention.
====
Ok. It seems too much for a comment already.
I hope something of that could be useful.
And I have some considerations and ideas that you missed or just might find interesting. They are based on personal experience of using several language learning apps and also on analysis of issues with Anki (reported and discussed on reddit and support forum).
====
1. There are a lot of people learning subjects other than languages, notably — medicine. They also need a good app with basically same requirements (SRS). Targeting this audience would extremely broaden community of users.
They, however, have some unusual demands regarding cards format and interaction (ordering items, image occlusion, matching pieces, etc).
My own concept is to move cards presentation/interaction completely to web-layer with API to core scheduler, and support creating cross-platform addons implementing any custom cards using web technologies.
2. It is crucial to make decks easily shareable. In spite of principle «your own cards work better» there exist usecases when people prefer to use prepared decks (notably «starter packs», or book courses converted), to collaborate on decks, or to provide new cards from teacher to students.
My idea is to separate card content from learning statistics, and to keep content as simple zip archives of yaml-based files along with all assets. This would allow to share decks just like normal files (without stats) and in non-obscure format, and also to use some CVS systems for advanced collaboration.
3. There is unimaginable variety of ways users want to import and modify their cards and collections. Most noticeable meta-usecase is importing from various other apps (like memrise and quizlet), csv files, emacs orgmode, etc. Occasionally some people need to do some batch processing like global search/replace or reorganize tags.
My idea is to use yaml-based format as a base for storing cards content. This would allow to use good old text editors to edit cards manually, and also to implement format-specific features in advanced editors like Vim, Emacs, Atom.
====
Also, regarding learning mechanics.
1. Quite obviously, any interaction works better than just recalling, because it involves substantial activity, and makes stronger neural response. The «tapping test» is wrong because it simplifies and actually reduces that activity (from typing to clicking).
However, the format of «word ordering» could be useful if used properly. Maybe some cases of pure grammar tasks (like Spanish direct/indirect pronouns, or German modals/infinitives and separable verbs), or in pure ordering tasks (like sorting aminoacids by polarization).
2. Many things are better to learn based in contrasting. Well known are minimal pairs for phonetic. The same is applicable to grammar, in form of choosing correct form. This way requires two interactive features: making choices and preview how answer «looks like» in context.
Selecting predefined choices is also considered a bad practice, but it still makes sense when task focus is to find some matching structures, and all possibilities might be not known to learner.
3. There's no need of negative response to errors like big red message with disgusting sound. It's quite possible to learn without any «stick» part.
4. Images intensify imagination and help to build more bright usage context. Even if they are not strictly related to definition. But they can easily be annoying when selected by other people. That idea of built-in image search by Gabe Wyner seems quite awesome, if implemented properly.
====
Not long ago, I have implemented Interactive cards for Anki, that demonstrate some of the features I wish to have in an app.
The cards allow interactions: typing, editing, reordering words, selecting options. User answers are usually visible in context before submitting. And there's no negative response — in case of wrong answer they show correct version and highlight it to attract attention.
====
Ok. It seems too much for a comment already.
I hope something of that could be useful.
Thank you for this comment, very helpful!
1) Web version of cards are mandatory and this is planned too.
For some specific areas I included «dictionaries» in Yarrow: Layout
2) AI can be helpful here, it can analyze students experience and offer a sequence of cards.
But for these algorithm statistics is required.
3) 99% users don't know what is Yaml/CSV and they afraid of word «Import». The most popular apps offer prepared cards and unfortunately this is the right way.
In Yarrow you can create your own dictionary and if words in it are popular the dictionary will be shared automatically.
===
1) Tapping test and Word ordering will be available in Yarrow but you cat switch them off. Some users need them anyway.
2) And again maybe AI will help here when the statistics is collected.
3) Yes, no waste of time.
4) I saw user images in Memrise and I usually did not have time to upload them :)
But the idea is good!
I will try your Anki cards, thank you!
1) Web version of cards are mandatory and this is planned too.
For some specific areas I included «dictionaries» in Yarrow: Layout
2) AI can be helpful here, it can analyze students experience and offer a sequence of cards.
But for these algorithm statistics is required.
3) 99% users don't know what is Yaml/CSV and they afraid of word «Import». The most popular apps offer prepared cards and unfortunately this is the right way.
In Yarrow you can create your own dictionary and if words in it are popular the dictionary will be shared automatically.
===
1) Tapping test and Word ordering will be available in Yarrow but you cat switch them off. Some users need them anyway.
2) And again maybe AI will help here when the statistics is collected.
3) Yes, no waste of time.
4) I saw user images in Memrise and I usually did not have time to upload them :)
But the idea is good!
I will try your Anki cards, thank you!
Yarrow is the best language learning application with the highest score of 400 points. The only problem is that it does not exist yet.Not funny.
The job is poorly done if you ask me. First, no language-specific apps, except for Skylang, which is on the 2nd (?!) place? No points on application prices. An app with 4.7 on GPlay at the bottom of the list… are you… sure? Raiting system looks inefficient. Where did all these numbers come from? «Without voices the application is useless (-100 points).» Don't mind your opinion, but that's not what I personally think. How about text then?) If an application is speech-oriented only, is it useless for language learning?
I'd advise to remove Yarrow and Skylang and rename the article to 'TOP-22 Multilingual learning apps', for the start.
Good day!
1) The most popular apps are not «language-specific». I can not analyze 500 apps because I do it for free.
2) Did you try Skyeng (not SkyLAng) yourself? It's paid, right, but it's good.
3) Drops is a very beautiful app and the rating is 4.7 but it's useless for learners. This is really interesting, right.
4) Price is not a metric of learning efficiency. I indicated really free apps as additional information.
I paid $12 per hour to online teachers, I can pay $30 per year for an app. And app will be more efficient.
5) All numbers (points) described in the article and in Google Sheets. I tried to make these points as objective as possible.
Excellent app must contain voices, texts etc or you need to install several apps and this will be less useful.
1) The most popular apps are not «language-specific». I can not analyze 500 apps because I do it for free.
2) Did you try Skyeng (not SkyLAng) yourself? It's paid, right, but it's good.
3) Drops is a very beautiful app and the rating is 4.7 but it's useless for learners. This is really interesting, right.
4) Price is not a metric of learning efficiency. I indicated really free apps as additional information.
I paid $12 per hour to online teachers, I can pay $30 per year for an app. And app will be more efficient.
5) All numbers (points) described in the article and in Google Sheets. I tried to make these points as objective as possible.
Excellent app must contain voices, texts etc or you need to install several apps and this will be less useful.
1) The most popular apps are not «language-specific». I can not analyze 500 apps because I do it for free.I'm just saying the good way is to analyze them separately. If you havn't analized the whole (at least, the major part) market, doesn't that mean there is no sense in your rating?
2) Did you try Skyeng (not SkyLAng) yourself? It's paid, right, but it's good.No. Because it's -infinity in my rate list. Know why? My study language is not english.
Excellent app must contain voices, texts etc or you need to install several apps and this will be less useful.An app may be able to handle only a single task, but if does that better than any other, it's an excellent app. What's more important, «Don't put all your eggs in one basket.»
Great research and features! Thanks for sharing!
Great post! People really need to see the difference, while there are plenty of apps. So sad you did not include Elevate, this app is mostly for advanced learners and native speakers, but if you have already reached this level and need to read a lot of specific literature, it would really help
Зарегистрируйтесь на Хабре, чтобы оставить комментарий
TOP-23 Language Learning Apps