Getting to Grammar: Comparing my method with Steve Kaufmann's method

Steve and I have been having a bit of an interesting back-and-forth on how to approach grammar. It started with my post entitled If you want accurate grammar quickly, Steve Kaufmann's method is not for you, and then moved over to the comments section of a tangentially related post on his blog.

The back-and-forth actually led me to think that we're a bit closer in our approaches than my earlier post suggested, so I thought I'd pull together all of the relevant comments into one place and then offer some further comparisons of our methods, making use of a few rough graphs. The fun begins, after the jump.

Read more...There's quite a chunk of text that Steve and I generated in the comments, so if you want to skip all that and get right to the (much briefer) comparison of our methods, click here. Otherwise, let's dive into the discussion.

First, you might want to take a look at the post that started it all.

Steve's reply came over on his blog:

I read your post and I am sorry but it makes no sense to me. I find that I cannot get the grammar rules into my head without first getting the exposure. … Input and vocabulary over grammar rules, anytime, at least in my experience.

Feeling like I hadn't quite explained myself well enough, I replied:

First I should probably be clearer about what it means to "get it in your head". I don't mean that you need to have it mastered, or that you need to have it memorized. I completely agree that that will most easily happen after lots of exposure. But anyone who picks up a grammar and reads that "estar" becomes "estoy" when "yo" is the subject already has it in their head to some degree, and that's the minimum you'd need to do to get it in your head. As I'm sure you're not saying that you can't get that in your head, I presume we're talking about different things.

The language in which I'm closest to the degree of accuracy that I seek is Japanese. … I would presume that your argument here would be that the input I'm getting is leading to my improved grammar, and I won't disagree that that's contributing, but whenever I come across something that doesn't fit into my understanding of how Japanese grammar works (extremely polite forms that I'm not accustomed to using come to mind), I dig until I understand how the rule works. That makes it much easier to understand it a second time. If I didn't get it the first time and didn't look it up, the second exposure would just be another puzzle.

And I'm definitely not saying that grammar is the only thing you focus on. When I've employed my approach, it typically takes me two or three weeks before I'm hardly spending any time on grammar at all. So you say "input and vocabulary over grammar rules", and I say "spend a few weeks getting the grammar into your head, then exposure and vocabulary over grammar rules".

… I think another point where you and I differ is that I think you should systematically try to fix your grammar, whereas my understanding of your approach is to just get it as it comes to you via input, filling in the gaps when you feel like it. (Correct me if that's wrong.)

Steve replied:

I constantly refer to grammar books, a little at the beginning and thereafter quite often reading the same rule or looking at the same table. I do not try to nail anything down. Eventually it all sticks. I have always said so. It is just that my point of emphasis is input.

I am actually quite accurate in my use of language, and constantly try to get more accurate. Expanding vocabulary is one of the best ways of achieving greater accuracy.

To which I replied:

I'm beginning to wonder if our positions on grammar are closer than I originally thought. Let me ask you a few questions:

1. When you start a language, how and to what extent do you review the grammar?
2. When you encounter grammar you don't understand, how and to what extent do you systematically try to figure it out? Every time? Sometimes?
3. What do you do when whatever you do in 2 is taking a very long time to stick in your head? (German cases are an instance of this for me; I needed (and, thanks to not using German frequently, now again need) to take steps beyond just getting exposed to them and reviewing my grammars to get them down.)

And, finally, Steve replied to my questions:

1) I have never started a language with LingQ. I have usually bought a little starter book Teach Yourself, or Colloquial and followed it, and then moved on to other content. I tend to skim the explanations and devour the little dialogues. I then refer back to the book from time to time. I do not try to understand or remember the explanations. In future, if the language is Dutch, or Czech or a language related to one I already know, I think I will just dive in, and only refer to grammar explanations later. If it is Turkish, I may still start with the starter book and CD and do LingQ in parallel.

2) Rarely look things up. It is usually when I am writing or speaking that I will look up a declension table of something. On the other hand, I will from time to time review a short grammar book, to see if things make more sense. I do not retain much, but I think it helps to make me more attentive.

3) When I was studying German on my own, I spent a fair amount of time trying to get the declension tables into my head. I was unsuccessful and now rely on having heard it so often and hope for the best. The same with Russian. I do save phrases in LingQ which feature the cases that cause trouble. I have tagged words and phrases for their case endings and reviewed them as a batch. I try to be attentive to these. I occasionally review the tables but find that the benefits are very short term. Mostly I try to notice them while listening and reading and hope for the best when I speak.

After Steve's last post, I decided to polish off a post explaining in some more detail my own grammar method, and there you'll find more details on the system I'm comparing to Steve's.

First, here's how time is spent on grammar under my method, as explained here:

And this is my understanding of how it's spent under Steve's method:

So Steve spends a little time at the beginning with some introductory materials. He then refers back to grammar as necessary, with exposure to the language (in particular, reading and listening) being his main source of exposure to grammar rules. He also engages in a similar refinement effort. However, because that is his main source of solidifying the rules outside of exposure, the amount of grammar time spent on that declines more slowly that it does under my method. Under both methods, it will eventually go down to next to nothing, but it will take longer under Steve's method to get to that point.

Accordingly, here's a rough comparison of the grammar understanding generated by the two methods (assuming that the average native speaker's knowledge of grammar is 100%):

Under both methods, an initial review of materials will quickly establish some basic understanding, but from there the level of understanding diverges.

Under my method, you get a large initial jump by outlining your grammar. This will cause a lot of it to stick. Not all of it, of course, but because you are actively processing the information, it is a much stronger form of exposure than the passive grammar reviews that Steve primarily relies on.

From there, both methods rely on a refinement process. Steve frequently consults grammars while I use my own refinement process only as necessary. Both processes result in a gradual approaching of a native speaker's grammar level, but because of the boost the outlining gave under my method, my method will get you there more quickly. Steve's method will of course get you there too, but it will just take more time.

In the post that started this whole discussion, I wrote that Steve's method won't lead you to the same level of accuracy in your grammar. However, given enough time and enough checking with grammar books, as Steve does, I think Steve's method can take you to at least an equal level of accuracy. The issue, again, is that I think this will take more time.

Thus, I think Steve's method will work if you've got both the time and the commitment to continuously get regular exposure to a language over a very long period of time. Many language learners can't or don't do that, so I would be wary about recommending Steve's approach to grammar. My approach on the other hand, will lead to a greater understanding more quickly, and can be put to good use when time is limited. After this comparison, I’m left with one question that I’d love to have answered: how much more time would Steve’s method take to reach accuracy parity with my method? Although I think the time difference is significant, I don’t have a good basis on which to make any guesstimates.

As a final note, and one that I’ve mentioned before when comparing my approach to Steve’s, I think our varying approaches very much reflect our varying goals in language learning. Steve main focus is to enjoy content in the language, while my aim is more centered on being able to use my languages at work, which means I need to try to get more accurate more quickly for output purposes than a nearly all-input method would allow.

Why output trumps input in language learning

OK, so I don't really think that output trumps input, but I thought I'd lead off with a contrarian title vis-à-vis Steve Kaufmann's post entitled Why input trumps output in language learning. Some amount of input necessarily needs to come before you can produce any output, but saying one trumps the other is like saying reading blogs trumps writing blogs; sure, you can learn a lot by reading blogs, but you'll only be getting your message out there once you start writing one. (And, incidentally, in either case, you'll be getting exposure to a language.)

The reason I went with a contrarian title was because, when I read Steve's post, I thought that most of his arguments for input learning could easily be changed to serve as arguments for getting into output sooner rather than later. Below I've edited Steve's post to show how easily those arguments can be turned in the other direction. I've tried to edit as little as possible. Some of the changes work better than others, and some even work surprisingly well, but they all go to my main point here, which is that early output is a good thing.

Read more... I've used red to mark text I deleted from Steve's post, while blue marks the text I added:

Some arguments in favor of output input. I am sure there are many more.

• We need to start speaking understand before we can speak well.
• I would rather communicate with people early understand well and stumble when I speak than communicate with people later and stumble less the reverse.
• If we can never pratice producing intelligible phrases and do not understand the answers, our conversations will not last long.
• Passive vocabulary is powerful, necessary, and always much larger than our active vocabulary of the words we like to use, so we need to start working on active vocabulary early and frequently.
• The more we can write and speak understand, and the more words we can use actively have, even passively, the more interesting our interaction with the language and the more words we can acquire.
• If we can actively use understand most of the words in a text or conversation, it is easier to pick up the words and phrases we do not yet know than if we merely understood everything passively.
• The ability to use active acquisition of passive vocabulary through output input, is like putting the pieces of the jig-saw together. Gradually the picture of what we're trying to express becomes clearer.
• Output Input is easy to arrange. We can speak listen and write read anywhere and anytime.

See here for some ways that the internet makes output possible from anywhere, which of course includes Steve's own LingQ.

• Output Input is interesting, if we choose content that is meaningful to us.
• If we develop the habit of producing output input learning, we become independent.
• Being able to produce output Input learning makes it easy to practice review our languages, and maintain them.
• Through producing output input learning, especially on topics we like writing and speaking about with authentic content, we learn not only the language, but many more things.
• At any time in our output producing input learning activities, we can decide to listen speak or read write, to practice what we can produce have learned.
• Of course we need to speak read a lot in order to speak well, but. Our progress in speaking will be smoother if we invest time in output input, and continue doing so.
• Our interaction with any language, including our own, is mostly as listeners and readers, so we need to make extra efforts to practice producing output.
• If we are good speakers listeners and writers readers, our output input skills will have a sound base.

One of my goals in any language learning project is to have little difficulty in conveying complex ideas to native speakers read a full length book in that language. Getting there is a powerful moment of achievement, an Everest.

I could go on....

A single workflow to make use of online language-learning tools

There are so many language-learning resources out there on the web, it's kind of tough to figure out how to make use of them all. In looking at how I'm using these tools myself, I put together the following little process to incorporate many of the language-learning tools I've been using into a single workflow:

Oh, and this workflow is completely free.

Let's walk through this, after the jump.

Read more... Start with reading and/or (but preferably and) listening to something in the target language. LingQ is all about content with both text and audio, so that's a good place to start looking, but you're hardly limited to LingQ; any recordings you can find with transcripts, unabridged audio books (including children's books), etc., will do the trick.

To the extent there's anything you don't understand in the text or audio, look it up and add it to your spaced-repetition system. Anki is my current SRS of choice, but some other popular choices are Smart.fm and Mnemosyne.

Then write something about what you read or listened to in the target language. Try to make use of whatever you needed to look up and add to your SRS, and to the extent that you need to look up anything else, add that to your SRS as well.

Then get that writing corrected. There are a number of ways to do this, but Lang-8 is my standing favorite, and italki recently implemented this feature. Again, if the corrections include things you need to look up, add them to your SRS system.

Once you've got the corrected text, record yourself speaking it and get that recording corrected by native speakers. I use Cinch and Lang-8 to accomplish this.

You've now written and read that writing. Now it's time for some plain old talking. Making use of everything you've learned thus far, record yourself saying something about the running theme and get that corrected in the same way you got the recording of your text corrected. Once again, if the corrections give you any thing that needs to go into your SRS, add it.

At this point, you should have everything you need to get in your SRS. Now go over to RhinoSpike and get native speakers to record the pronunciation of each of those words. Take those audio recordings and add them to your SRS system. From there, you just need to review your newly added items as part of your regular SRS review.

You've also got two things that you've recorded yourself: your corrected text and some plain old talking. Go to RhinoSpike again and get a recording of both from native speakers. Once you've got those recordings, add them to a playlist on iTunes and listen regularly. I'd recommend just throwing all of these recordings into a random-order playlist and listening to them in the background while doing other things. This will provide a review of all of the above.

This entire workflow can be tailored to your level. At the most basic level, you can even use children books; my kids have plenty of books that come with audio CDs in all three of their languages. But you don't necessarily need to dumb the text down; you can also just keep it short. For example, if you're just starting a language but want to read a news article, you could limit yourself to just the first paragraph. This will likely take a while, but it won't be insurmountable.

If you've got a way to make this workflow, I'd love to hear it!

Get audio recordings of any foreign language text for free

Earlier, I told you about how to get foreign-language text read to you online for free through computer-generated text-to-speech software. Text-to-speech software still doesn't quite pull off completely native language, so wouldn't it be great if you could actually get a native speaker to record some audio for you?

That's the whole concept behind RhinoSpike, a new, completely free website launched on Thursday by Thomas Hjelm and Peter Carroll, the two guys behind the language-learning blog Babelhut.

Read more... Here's how Thomas described RhinoSpike via email:

You submit [the target language] text you want to be read aloud/recorded by a native speaker. It goes into a queue for that language. Native speakers see your request, record their voice and upload the audio file. You download it and add it to your Anki/SRS flashcards or load it onto your MP3 player or do whatever you want to do with it.

You can also record your voice for people learning your native language. Doing so bumps your own requests forward in the queue, so native speakers will see them faster. Help others and you receive help in turn.

Helping others and getting bumped to the front of the line is a nice touch.

Thomas went on to explain how this might be used in conjunction with another of my favorite language-learning sites:

You could think of it like Lang-8 for audio files, except instead of getting corrections you are getting audio for any text you want. In fact, you can use the two sites together. Write a journal entry on Lang-8 and get it corrected by native speakers. Post the corrected journal entry on RhinoSpike and get it read aloud for you by a native speaker. Use the audio file for listening or speaking practice.

This meshes quite nicely with being able to get your own spoken language corrected on Lang-8, but I'd do it a little differently than Thomas suggests. After getting your writing corrected on Lang-8, submit your own audio recording of the text as an entry on Lang-8 with a link to the text on RhinoSpike. Then Lang-8 users can tell you what you're doing wrong on Lang-8 and provide you with a correct recording on RhinoSpike. And I don't think Anki and Lang-8 are the only tools that RhinoSpike will find synergies with. LingQ, for example, is all about having audio paired with text.

For RhinoSpike to be good at what it's trying to do, it'll need to obtain a critical mass of users. Given that it was launched just two days ago, it's nowhere near that point. As of this writing, there aren't more than 80 members on the entire site (4 pages in the profile list, a max of 20 profiles per page).

Given the number of users, it's not surprising that there aren't that many requests for recordings up there yet. Before I added some stuff to the site, there were only 17 audio requests on the audio request page, with Japanese topping off the list with 7 requests. And, of those 17 requests, there were only 6 recordings, and all were in Japanese done by a single user. I recorded four more in English and had my wife do one in Japanese, which made a total of 11 audio recordings in English and Japanese done by two users. Assuming that kind of participation rate is typical, their user numbers need to go way up to make this the kind of tool it has the potential to become.

After putting up my own recordings, I posted a bunch of requests for recordings of texts in six different languages. The same user who did all the other Japanese recordings came almost immediately and fulfilled my Japanese requests as well. The others (Chinese, Spanish, Portuguese, French, and Italian) remain unfulfilled. With more users, RhinoSpike could very well become like Lang-8, on which native-speaker input is often immediate but in any case never takes long.

One thing that's probably holding back users on RhinoSpike from adding recordings is that it's a hassle to put the recordings up there. You've gotta record your own audio file and then upload it. They recommend Audacity, but I found it easier to simply record my voice with the Voice Memos app on my iPhone, sync the iPhone with iTunes so that the recording ends up in iTunes' music list, right click on the track in iTunes and convert it from an M4A to an MP3 from the contextual menu, dump the MP3 on my desktop, and then upload the MP3 to RhinoSpike. Recording from directly within the RhinoSpike web app, a feature found on Livemocha, is coming in the next version, but for now anyone who wants to upload audio recordings has to go through the hassle of using some other app to generate the audio file. I don't expect that many users will go through all this work to put up audio recordings.

Another issue that's going to limit users is the number of language localizations. It's currently available in English, Spanish, and Japanese, while it's possible to submit requests in a ton of other languages. I doubt the website will see nearly as many native-speaker members in languages for which it is not localized, so hopefully they'll start the crowdsourcing efforts to localize for various languages, as is common on many language-learning websites.

There are two other things that I noticed that could be improved to make the website easier to use. First, searching for friends is a pain. There's no way to filter the profile list to find native speakers of the language you're learning (and searching for, e.g., "Japanese" or "English" strangely produces no results at all). Second, there's no way to quickly find recordings. I'd love to be able to quickly look at all available recordings in a given language to be able to hear native speech, but there's no easy way to do this. As is, you've gotta click on the request and then, if there's a recording (and that's still a big if), you can listen to it.

Nevertheless, I would pretty much chalk off all of the above to the site still being a just-released web app. The bottom line is that RhinoSpike is a great start for a language-learning tool with a lot of potential, and I hope to do my (self-serving) part in bringing more users to it.

However, there is one thing I still don't get... how the heck did they come up with the name "RhinoSpike"?

NY Times' "The Web Way to Learn a Language" is misleading and incomplete

The New York Times last week ran an article by Eric Taub entitled "The Web Way to Learn a Language". For the most part, the article is an uncontroversial list of some of the better known language-learning resources on the web, followed by a grab bag of a few lesser-known, language-specific resources plus a few iPhone apps.

That the article is an incomplete list of the numerous resources available on the internet is probably the nature of the medium, but it also to some extent reveals Eric's prejudices about language learning: that some kind of structured "class" is needed, along the lines of those found in the offerings of Rosetta Stone, TellMeMore, Livemocha (see my review of it here), Babbel, and BBC Language. Some things that are not really part of a course fall into his grab bag at the end, but he completely misses out on great resources like iTalki, Lang-8, or LingQ, which respectively can be used, among other things, to let language learners freely tackle whatever content they like in speaking, writing, and reading and listening.

However, the article is shockingly misleading in how it characterizes the results of one language learner's experience.

Read more... Here's what the article says:

The young woman … was born in Iran and spoke only Farsi until her arrival [in the U.S.] two years ago. What classes, we wondered, had she attended to learn the language so well?

There's that assumption that a "class" is needed, plain as day.

"I didn't," she said. "I used RosettaStone."

And that's where the article leaves it. And what are you left thinking after that? Naturally you end up thinking that Rosetta Stone is the only thing you need to sound just like a native speaker. But let's rewind and repeat for a second…

…spoke only Farsi until her arrival [in the U.S.] two years ago.

Uh… say again? She's been living in a place where she's getting tons of exposure to her target language for TWO YEARS?

For those of you who are wondering, living in the place where your target language is spoken will generally do wonders for your language abilities. Let's assume she speaks Farsi at home. I would still wager that she's been going to a U.S. school, has native-speaker friends, watches U.S. television, reads U.S. websites, has an English-language Facebook account, etc. To slavishly suggest what the marketers are hoping would be suggested—glory be to the software!—without checking to see what other exposure she might have been getting to her target language is practically negligent.

Indeed, the sole fact that she's been living in the U.S. for two years could be more than enough to explain her native-sounding English. A friend of mine from Belarus moved to the U.S. when she was 15. I met her when she was 18, by which time she was completely indistinguishable from a native-English speaker. After spending about two years in the U.S., my wife began getting asked if she was a native-English speaker. After just a year in Japan, even I was able to briefly fool people on the phone into thinking I was a native-Japanese speaker. And none of us had used any software, while all of us had spent has spent significant time in places where our target language was spoken.

If the article's young woman had just arrived in the airport from Iran speaking native English and said the only exposure she had to English was Rosetta Stone, then I would be very impressed indeed. But to uncritically suggest that exposure to English via Rosetta Stone's software somehow played a more prominent role in her language learning than other avenues of exposure—especially for someone who in all likelihood was getting a lot of exposure to her target language—is doing readers a disservice.

Link: The Web Way to Learn a Language [New York Times]

Why didn't my university teach languages like Drake University does?

We in the language-learning blogosphere are generally not impressed by university-level language programs. Some of us have even gone so far as to envision a brave new world of institutional language learning where entire language departments get the boot and students take advantage of native speakers, study abroad, and the multitude of resources available to them to learn their language of choice.

Well, I hate to spoil our "We know so much better than crusty, old schools" party, but Drake University, "a private, fully accredited, coeducational university on a 120-acre campus in Des Moines, Iowa", seems to be way ahead of the curve on this one. They implemented just such a system. And they did it in 2001. To those of you with short memories, they launched this way back when you couldn't watch foreign-language videos on YouTube or listen to language-learning podcasts on your iPod because, well, when it launched, YouTube, podcasts, and even the iPod didn't exist.

So what exactly has Drake been doing since they jettisoned their language faculty? Read more... Here are the outlines of their approach.

• At the beginning of their language studies at Drake, students take a course on language-learning strategies in English that is not aimed at any particular language (sounds like the book we're working on).


• Students meet three times per week in groups of no more than four with a native speaker of their target language and speak nothing but the target language during that time (sounds like LingQ's group sessions). Some classes are now completely virtual, via Adobe Acrobat Connect and Skype, making it seem even more like LingQ.


• Outside of these meeting times, students "practice using the language, make audio recordings of themselves speaking, and complete a variety of other assignments as part of the required electronic portfolio", which includes a journal in the target language (like Lang-8), the aforementioned recordings (as can be done on Lang-8 or Livemocha), writing samples (as can be done on a bunch of language-learning websites), and other things.


• Over the semester, students meet with a Ph.d.-holding linguist to cover grammar questions in English, go over how they're doing, etc. The linguist's main role seems to be a coordinating one.

Drake's method seems to be spreading slowly, with some schools adding additional advancements. Inside Higher Ed describes the case of Abilene Christian University:

Abilene Christian piloted Mandarin during the 2008-9 academic year using the Drake model of a supervising professor and a native speaker conversation partner. The professor … was in Beijing, and on-campus graduate students fluent in Mandarin led discussions. Arabic is taught by a professor in Tunisia.

Now that technologies like Skype are so commonplace, native-speaker teachers who live in their native countries seems like such a no-brainer to me.

And, most importantly, the model seems to be working. According to Inside Higher Ed:

There has been no comprehensive study of how Drake’s students compare to students who learn languages in a more traditional way. But the anecdotal evidence is there, many times over, said Jan Marston, director of [Drake's program] from its founding until last year.

When students trained at the Des Moines, Iowa, university study abroad, she said, “they’re placed in classes way above where the seat time would indicate they should be.” Students report back that while other students in their programs abroad speak English to each other, “Drake students are speaking Russian to the Russians.”

Marc Cadd, who directs Drake’s [program currently] said students are generally placed two semesters ahead of where they would be at Drake when they study elsewhere. For instance, students who had finished Drake’s Spanish 101 and 102 classes would likely be placed into a third-year language class when studying abroad in a Spanish-speaking country “primarily on the strength of their speaking skills."

I can't say I'm surprised. The approach they're taking jives much more with what I've found in my own experience than any more traditional approach.

So, Drake University, my hat's off to you. Your program is by far closer to how I would have liked to have learned languages in college, and your results certainly do seem to show it. (And someone might want to tell Steve Kaufmann to give these guys a call, given just how similar their system is to LingQ's.)

Links:
Outsourcing Language Learning [Inside Higher Ed]
Languages without Language Faculty [Inside Higher Ed]
World Languages and Cultures [Drake University]

Help me with Steve Kaufmann's Wikipedia page!!

After hearing about Steve's run-in with a little Napolean over on Wikipedia, I thought I'd insert myself into all this fun, so I started a Wikipedia page on Steve himself. I've gone on there a few times before to check out Steve's bio, only to find myself surprised that no one had put anything up yet.

The same is actually true of LingQ as well. I do think that if Livemocha and Lang-8 get Wikipedia pages, then there's no reason why LingQ shouldn't have one. However, let's let that one cool off for a little bit and focus on Steve's entry for now.

Here's how I started it:

Steve Kaufmann is a Canadian polyglot linguist, author, award-winning blogger and the founder of the language-learning website LingQ. He currently speaks twelve languages to varying degrees of fluency: Cantonese, English, French, German, Italian, Japanese, Korean, Mandarin, Portuguese, Russian, Spanish and Swedish.

I modeled the text after Michel Thomas' Wikipedia entry. I made lots of citations, but as the article's not so long, I marked it as a stub in the hope that you guys would step in and expand it.

To keep this from getting deleted, remember to cite! cite! cite! Steve's book is up on the web completely for free, and it's full of good, citable information (Wikipedia loves citations to books). And feel free to dig up any information that might be floating around the internets, especially on official sounding stuff (wasn't there an NPR interview a while back?).

Also, Steve, you're not allowed to edit your own entry, so please don't! But if you've got links to media coverage, that'd be helpful. And, of course, if "anonymous" comes along and edits the entry, hey, who's the wiser?

Get your foreign-language writing corrected online for free

When you're learning how to write in a language, there's nothing quite like getting your writing corrected. And when you're getting it corrected, there's nothing quite like getting it corrected totally for free. And when you're getting it corrected totally for free, there's nothing quite like getting it corrected for free and quickly.

Sound like something you'd be interested in? A comparison of the websites on which you can do just that, after the jump.

Read more... The first two sites below—Lang-8 and CorrectMyText—are focused primarily on textual corrections. The rest—Livemocha, Busuu, and LingQ—include textual corrections as one among many features.

One note before diving in... the comparisons below are only looking at what these websites do in terms of text corrections. All of these sites can of course do other things, but I'm leaving those features aside for now (although feel free to highlight your favorite features in the comments below).

Lang-8

• Overview. Lang-8, based in Tokyo, is a two-person project by Yangyang Xi, CEO, and Kazuki Matsumoto, CTO, that focuses letting language learners get their texts corrected.


• Content. Lang-8 is set up as a journal or a blog, but you're free to post whatever text you feel like posting. Although many people do post journal-like entries, I typically post all sorts of things in there. In addition to texts to get corrected, this mainly consists of language-related questions. Just as people are happy to correct your text, they're also happy to answer questions about whatever confusing point of the language you've come across.


• Making corrections. Lang-8 first breaks the text down into sentences, separating them based on punctuation (this results in the occasional weird break-up when you have something like "12.1" in the sentence; Lang-8 interprets the decimal point in that number as the end of a sentence and breaks it up accordingly). Then correctors can edit sentence by sentence. The system flags uncorrected sentences so subsequent correctors can focus their efforts where most needed.
  
  Correctors edit each sentence in a little window. The one annoying thing about the editing process is that, if you want to add formatting to the text, you've gotta deal with tags tossed into the text in that little window, such as [BLUE][/BLUE] or [BOLD][/BOLD]. It can get pretty jumbled up.


• Speed of corrections. Although none of these sites are slow in getting corrections back to you, the corrections come extremely rapidly on Lang-8; I rarely wait an hour, but I think the most I've ever waited is something like a day. In fact, one day I put up a whole bunch of posts on Lang-8 and, by the time I was done adding all the posts, most of them had already been corrected.


• Correction presentation. It is up to individual correctors to make their changes apparent through formatting: bold, strike-thru, red, and blue text. Your results will vary, but most correctors do a good job of making it easy to see what they've changed.


• Languages. You can post in any language you want, and native speakers of all major languages are well represented on the site. I make most use of Japanese, unsurprisingly, but I've also made use of Chinese, Spanish, Portuguese, and French thus far. I'd wager that it'd take longer to get corrections for less frequently studied languages, but I've not tested that hypothesis.


• Interface. Lang-8's interface is alright; it's nothing to rave about, but it gets the job done. I'd like them to make it even easier to view edits, but it's decent as is.


• Bottom line. I find Lang-8 to be the best of the bunch, and I recommend it highly.

CorrectMyText.com

• Overview. CorrectMyText, based in Russia, is the project of Dmitry Lopatin. It's a new entry to the free online text-correction market; as far as I can tell, it was launched all of seven days ago. As such, it's still got a lot of squeaky wheels that need some grease, but the functionality you need to get text corrected is already there.


• Content. You can put any kind of textual content into CorrectMyText.com.


• Making corrections. CorrectMyText first breaks the text down into paragraphs, separating them based on line breaks. The corrector can then edit each paragraph's text direcly.


• Speed of corrections. Given how new CorrectMyText is, and thus the limited number of users it has compared to the other sites in this list, the corrections don't come quite as quickly. Nevertheless, if my limited experience is representative, you'll still get them within a day or two.


• Correction presentation. The corrector cannot apply any formatting. CorrectMyText.com will automatically create side-by-side before-and-after versions of the text. The before version will show the edited text highlighted in red and struck through. The after version will show the edited text highlighted in yellow. The learner then has to compare correction by correction to see the changes.


• Languages. Chinese, English, French, German, Italian, Japanese, Korean, Portuguese, Russian, Spanish.


• Interface. This is still a bit rough. It's sometimes hard to figure out what you need to press to move on, and I found myself pressing the wrong thing more than once. It remains very basic, as you'd expect from a newly launched website.


• Bottom line. As a new entrant to the market, it still needs some work before it'll be a viable contender against Lang-8, but it's definitely a site to keep an eye on.

Livemocha

• Overview. Livemocha's main product is it's Rosetta Stone-like language-learning courses, but the coolest thing it does is connect you with tons of native speakers, including through text corrections (see my complete review of Livemocha here).


• Content. The textual submissions on Livemocha are at least nominally supposed to be based on prompts connected to lessons, e.g., "Describe the locations of a set of people and objects". However, there's nothing to stop you from writing about whatever you care to write about, and indeed that's what I've often done. In fact, Livemocha may soon be considering implementing freestyle writing. That'll be more than a nod to reality than an actual change, but I'd be happy to see the addition.


• Making corrections. Correctors simply get a comment field in which they can make comments and variously format the comment text.


• Speed of corrections. Livemocha has a very large user base, so corrections come back very quickly, certainly comparable with Lang-8.


• Correction presentation. Like Lang-8, it is up to individual correctors to make their changes apparent through the various formatting options that are available. Again, your results will vary, but most correctors do a good job of making it easy to see what they've changed.


• Languages. Arabic, Bulgarian, Chinese, Czech, Dutch, English, Estonian, Farsi, Finnish, French, German, Greek, Hindi, Hungarian, Icelandic, Italian, Japanese, Korean, Polish, Portuguese (Brazil), Portuguese (Portugal), Romanian, Russian, Spanish, Turkish, Ukrainian, Urdu.


• Interface. As far as text correcting goes, I've got no major complaints. The interface allows you to get the job done.


• Bottom line. Not a bad back-up to Lang-8 for text corrections, but as Lang-8 specializes in this feature and it's just another feature at Livemocha—and Livemocha's still not made for freestyle writing—I'm going to stick with the specialist Lang-8 and hope that Livemocha gives this feature some TLC.

Busuu

• Overview. Busuu is a direct competitor of Livemocha, using a similar picture-based learning method, but it also connects you with lots of native speakers, including, again, through text corrections.


• Content. Just like Livemocha, the textual submissions are at least nominally supposed to be based on prompts connected to lessons, e.g., "Describe a real person in your life", but, again, there's nothing to stop you from writing about whatever you care to write about.


• Making corrections. Correctors simply get a comment field in which they can make comments and variously format the comment text, mirroring Livemocha. It does have one convenient feature that Livemocha lacks: a button to automatically copy and paste the unedited text into the comment field.


• Speed of corrections. Although I don't have any numbers to back up my supposition, it seems to me that Busuu has less users than Livemocha, and accordingly will take a little longer. That said, corrections still come back within a day or so.


• Correction presentation. Like Lang-8 and Livemocha, it is up to individual correctors to make their changes apparent through the various formatting options that are available. Again, your results will vary, but most correctors do a good job of making it easy to see what they've changed.


• Languages. English, French, German, and Spanish. One of the largest differences with Livemocha is that Busuu covers fewer languages.


• Interface. Busuu's interface is probably the nicest of the bunch, and it's just fine for getting texts corrected.


• Bottom line. Given how similar it is to Livemocha, the bottom line for both is essentially the same; not a bad back-up to Lang-8, but until Busuu puts some more focus into textual corrections, I'll be sticking with Lang-8.

LingQ

• Overview. LingQ's focus is on audio and textual content (especially audio with the accompanying textual content), and, among other things, it has a feature that allows you to get your text submissions corrected. LingQ's text correction feature, however, is not free (it's not terribly expensive though, basically coming down to $0.033 per word, although the pricing is a bit more complex than that). I've broken the free-stuff-only rule and included it here because it has some very interesting features that the completely free ones do not yet match.


• Content. You can put any kind of textual content into LingQ.


• Making corrections. You highlight the text you want to correct, and click a button. Up pops a window with the text you selected, and you can then edit it. Thus far, that pretty much makes it like all the rest. But then you then get the option to select what kind of error it is—spelling, word order, verb form, etc.—and that data will be used when presenting corrections.


• Speed of corrections. Corrections are generally done by a learner's selected tutor, and you might have to wait a little bit before your tutor has a chance to correct your text. That said, tutors seem to reply relatively quickly. I'm a tutor on the site, and I typically try to do my corrections as soon as I'm notified they're there. My slowest response time thus far has been a single day.


• Correction presentation. Just like CorrectMyText, LingQ will automatically create side-by-side before-and-after versions of the text. The before version will show the edited text highlighted in yellow, the after version in green. The learner then has to compare correction by correction to see the changes. Alternatively, the same corrections are listed out below the side-by-side versions in a table that also lists correction-specific notes and the type of each correction.
  
  And then here's where LingQ lays down some awesome. Using the type of errors that the corrector marked down, you get an analysis of your mistakes.
  
  
  Just. Fricking. Awesome. Getting this level of analysis is far better than just seeing your mistakes, because it can help you focus your efforts on where to improve. Although Steve at LingQ is not a big fan of focusing on grammar, this lets you do just that. If you see that you're struggling in a particular place, you can do a read-through of the section in your grammar on that topic, or take other steps to figure out why you keep messing up. Great feature.


• Languages. Chinese, Dutch, English, French, German, Italian, Japanese, Korean, Portuguese, Russian, Spanish, Swedish.


• Interface. LingQ recently made some system-wide improvements to the site, which included some interface improvements. The site before was fine, and the improvements made it better. Overall, a very usable interface.


• Bottom line. They've built in some very clever features into LingQ's textual correction system, but I just can't justify the cost for text corrections when Lang-8 and all the above are available completely free of charge.

So do you know of any other places where we can get our foreign-language writing corrected? If so, drop a line in the comments!

Online students outperform those receiving face-to-face instruction

A video entitled Social Media Revolution has been making the rounds via, ahem, social media. The video makes the case that social media is "the biggest shift since the industrial revolution".

The video, which tosses out a couple of interesting things for language learning, after the jump.

Read more... Probably the most interesting fact they toss out in respect of language learning is this (at 1:00 in the video):

2009 US Department of Education study revealed that on average, online students out performed those receiving face-to-face instruction... 1 in 6 higher education students are enrolled in online curriculum

So, if you've run into a language teacher who's skeptical about your use of eduFire, Lang-8, LingQ, Livemocha, or any of the rest, hold your ground because you've got some good statistics on your side.

At 2:00 in the video, it points out a language-learning resource that has certainly not gone unnoticed by language learners:

Wikipedia has over 13 million articles. Studies show it's more accurate than Encyclopedia Britannica. 78% of these articles are non-English.

That means there's a pretty darn good chance that you can get materials in your target language on Wikipedia (and, of course, for you English learners, that means that 22% of the articles on there are in your target language).

With no further ado, here's the video:

List of content with both audio and text, courtesy of LingQ

LingQ is maintaining a great list of foreign-language content sources here (it's so good I also added it to the links in the right-hand column of this blog). What makes it so great?

This list is for users of LingQ.com to identify sources of language content consisting of both audio and text.

Listening to audio is great, but it's even better when you can refer to a text to find that word you didn't quite get.

I'm currently working on setting up a bunch of podcasts that I can regularly listen to so that I can keep languages I'm not using so much moving forward, and this list has some good leads for me. I've been able to dredge up some thing that aren't on this list, so I'll be dropping LingQ a line to add them in. If you know of any similar content sources that are not yet on the list, I encourage you to do the same.

Links: LingQ

Tutoring on LingQ

When I saw this post on Steve Kaufmann's The Linguist, calling for tutors on LingQ, I jumped at the opportunity. I've been enjoying the "you scratch my back, I'll scratch yours" tutoring on sites like Livemocha and Live-8, so I wanted to give it a go on Steve's system as well. Tutors on LingQ get points and can cash out those points at $15/hour or use them on LingQ, and my points will surely be used to further my own language-learning goals. (Steve used LingQ to learn Russian. Perhaps I should finally make an attempt to take Russian off of my list of unfinished business...)

While LingQ will certainly be up for a more thorough review by me after working with it some more, I can say right now that there's one feature I absolutely love. On LingQ, you copy and paste any text you want into it. When you highlight a word in that text and click a button, LingQ will look up the word for you automatically using Babylon's dictionaries or other free resources like Wikipedia, and then you can quickly make a flashcard (or what's called a "LingQ" in LingQ) by simply cutting and pasting. What a blessing that system is. For years, I've been taking my arbitrary texts (whether news articles, lyrics, or what have you) highlighting all the words I didn't know, looking up all the words, and then making flashcards. LingQ makes this exercise so much easier. The only downside is that you're limited to 300 flashcards on a free account, but if you can shell out (a pretty darn reasonable) $10/month, you'll have unlimited flashcards.

To get back to my own LingQ tutoring, I'll be holding my first session next Monday on a topic everyone seems to be wagging their tongues about: this big, bad economic crisis. So, if you're studying English, feel free to get on there and look me up!

Update Jan 15 2009 8:43PM: My username on LingQ is VincentPace (thanks Edwin!).

Related: Livemocha review: Love the native speakers, the method not so much

Livemocha review: Love the native speakers, the method not so much

I've recently been giving the totally free language-learning website Livemocha a spin. Livemocha is absolutely excellent for putting you in touch with native speakers and having them correct your written and spoken submissions, but its teaching method leaves a lot to be desired, and they still have some kinks to work out of the system.

Livemocha divides a language into courses, then units, and then lessons. For most languages, there are four courses that aim to get you to an intermediate level, and each course is divided into three units of about five lessons each. Lessons, in turn, are divided into four types of activities: learn, review, write, and speak.

Let me start with the last two and what I love about the site: how it links you up with native speaker tutors, and plenty of them at that. The "write" section asks you to write a short text, generally based on the lesson but you're free to meander off topic (and I frequently do), and the "speak" section asks you to read and record a passage of target language text. You then submit these to up to ten other users to correct for you.

Read more...Ideally, you'll want to submit your work to be corrected by native speakers, but even among native speakers your feedback will vary greatly. I initially began by just randomly selecting German speakers from among my friends and the website-suggested users, but I was able to quickly discover and prefer those who were giving me the highest-quality feedback. I now have a core group of tutors to whom I consistently submit such assignments to, and their feedback is phenomenal. They drill into my work to find even subtle mistakes and offer excellent explanations of what I'm doing wrong. So, while initially you may find that the feedback you get is not all that great, as you separate the wheat from the chaff you'll eventually end up with excellent tutors.

The other way in which Livemocha connects you to native speakers is via chat. You can do text chat, audio chat, and video chat. Livemocha encourages you to chat via their system by providing you with points for using it (more on that below), but given the rough feel of their chat capabilities I often find that we end up taking it out of Livemocha to MSN for text chat and Skype for audio or video chat. Despite the issues with Livemocha's own chat features, it stands as an excellent tool for putting you in touch with native speakers of your target language.

And you might be wondering how it is these people will correct your work for free. Like certain other other language websites, they use a deviously clever all-carrot, no-stick point system. You get points for studying, but also for teaching, i.e., doing things like correct others' written work. After you submit in the writing or speaking sections, you're presented with another learner's work to be corrected in your own language. This is ingenious social engineering; right after you've asked a bunch of people to correct your work, you're presented another's work to correct. How can you not? You actually can skip it, but I'd bet the skipping rate is pretty low.

You'll also find that, once you have your established tutors in your target language, you'll be eager to correct any work they send you in a quid pro quo; they're doing a great job for you, so you feel the need to do a great job for them. My only gripe, and I suppose it's more of a request for improvement than a gripe, is that I'd like to be able to sort my requests for corrections by the number of times the sender has corrected my work. For now, I do it manually by just trying to remember who has been helping me out.

Now let's turn to the parts that don't impress me so much. The "learn" section of a unit consists of a picture being shown with the text describing that picture below and a native speaker speaking the text. It's not always clear what the text is describing, so you're provided with a translation button that lets you see what the text is supposed to say in your native language.

The "review" section consists of exercises, of which there are three types.

1. Read: You select the picture that matches target language text.
2. Listen: You select the picture that matches target language audio.
3. Magnet: You put together a sentence magnet puzzle to match target language text or audio, which looks like this in the case of text:

Additionally, there are extra optional exercises, which include the above three plus "quiz" exercises, in which you are presented with a word, phrase, or sentence in the target language and must select the corresponding translation.

You can also make flashcards from the content in the lessons, or you can make your own flashcards from scratch. It's something of a hassle to make flashcards, and the testing method is the same as the "quiz" exercises, with incorrect answers selected from within the same flashcard set. These basic flashcards seem like something of an afterthought, and it's quite a hassle of pointing and clicking to make your own flashcards.

The core method is much like Rosetta Stone's; they provide you with the language, and you're supposed to figure out the rules.

As is always the case with such inductive systems, the problem is that that does not work very well for anything above a certain degree of complexity. I've been trying out Livemocha as a way to review German, and one of the issues I knew that I definitely need to review was the cases. The one-line explanation of German cases for the uninitiated is that certain German words, including nouns, adjectives, "the", "a", etc., change their form depending on how and after what they are used in the sentence. I had cases down pat before, but as I've not been using German a lot over the past few years the exact rules have slowly leaked from my head, and I thought I'd be able to pick them up using Livemocha.

But that was not the case. I frustratingly found myself making the same mistakes over and over again, and wishing I just had the rules presented to me so I could quickly refresh my memory. Ultimately, I turned to other websites and some grammar books I have to get a refresher. If this is the case for me, a person who is reviewing the rules, it would only be that much harder for someone taking their first crack at German to actually figure out what is going on in the grammar just by going through Livemocha's courses.

And I'm certainly unimpressed with the exercises' ability to actually test your knowledge. For one, you can often figure out the answer from words you learned earlier without needing to test the words in the most recent lesson. For instance, if you're studying adjectives, they might have "a fat man", "a skinny girl", "a tall boy", etc. But because they use a different noun for each, you can easily figure out what the answer is without knowing a thing about the adjective. Similarly, you can often use process of elimination to figure out answers, without really needing to understand what's being presented to you. For instance, pictures are often tested in groups of four. Once you've done the first three, you know the next answer will be the fourth. The same sort of process of elimination can be used in the magnet exercises. What's more, in the magnet activities, there is no tolerance for incorrect punctuation or the like. For instance, you might find "gut" and "gut." (i.e., one with a period and one without) as two separate magnets among the options. If you accidentally put the one without a period at the end of a sentence, it'll mark it wrong. While strictness has its place, this is most likely just a stupid mistake that doesn't reflect on your comprehension and hence should be ignored, but isn't.

Another practice I find suboptimal is their use of a single learning course for multiple languages. There is a core course that is simply translated to other languages to expand the system. While this makes it easier to incorporate more and more languages, it is not optimal for learning as the course will undoubtedly work better with some languages than others.

The last big group of issues with the site that I'll touch on are what appear to be growing pains: kinks that I would hope are temporary and will be worked out over time. These are the little things that take away from the experience.

Certain assignments ask you for things that haven't been taught yet. For instance, in German 101, Unit 2, Lesson 2, the writing assignment is "Describe the locations of a set of people and objects. Describe each. EX. The woman is on the yellow couch. She is not in the brown chair." However, up to this point the course has not covered how adjectives change in front of nouns. This means that your poor reviewers will have to correct all of your guesswork and it greatly increases the burden on them.

And the system still has mistakes outright in it. For instance, in one exercise, I came across this picture:



The text for this was "Wo ist er? Es ist im Karton." ("Where is he? It is in the box."). This is, of course, as wrong in German as it is in English, but it was that way in both the text and in the native speaker's recording. You would think that the native speaker would have at least flagged this for them so they could fix it instead of just reading it rote (if that was in fact a computer's voice, color me impressed). Another mistake I came across was "Der Junge hat keine roten Haaren" ("The boy doesn't have red hair"). The mistake is that there's no -n on the end of the word for "hair"; it should be Haare.

In addition to outright mistakes, there are also times when two or more pictures are the right answer, leaving you guessing blindly as to which one is actually the "right" answer. In the exercise below, the text says, "Where are they? They are in the box," and you've got to pick the correct picture. Well, are they referring to the candies in the box or the flowers in the box? It's totally unclear and you're left guessing which is supposed to be the right answer.





And here's another one. The text says "She doesn't have red hair." We can eliminate the guy and the lady with red hair, but which of the two non-redheads is this referring to? Only haphazard guessing will tell.





There is also generally bugginess in the responsiveness and behavior of the interface. There were a few times when I went through one "review" section and only got one or two wrong (out of 40) and ended up with a score like 70%. I can only attribute that to some sort of technical screw-up. There were at times time lags that resulted in incorrect clicking, and sometimes a click wouldn't register at all. This is particularly true when you have a Livemocha chat window open and are getting a new chat message.

Despite what now looks like a post full of griping and moaning, I would recommend Livemocha as a tool for language learners. Their teaching method is not all that great, but it's not terribly painful to click through a bunch of cards, and it's certainly helpful to hear the target language spoken by a native speaker. And, of course, you can skip it, if you want to. But the real gold lies in the site's ability to put you in touch with native speakers, and you should definitely arm yourself with that as one tool in your language-learning kit.

Pro bono language teaching

This article on Unlimited Potential, an Arizonan group that helps immigrants to the US (presumably mostly Spanish speakers) learn English, funded primarily by grants, describes how they are making a difference in their community through language education. As a lawyer, I'm used to the idea of spending a certain portion of my time working pro bono providing legal services to those who otherwise could not afford them and it is interesting to see how the concept is implemented in other industries, and the language-learning industry in particular.

With resources available online, it seems to me that it would be trivially easy to enhance language education to groups such as those targeted by Unlimited Potential simply by providing them to access to computers with an internet connection and a webcam. You could easily get them on websites like Livemocha.com where they can learn for free with the help of native speakers. Public libraries seem like an obvious resource for this, with the only problem being the silence typically demanded by libraries. At my local library, for instance, computers are sitting in a sort of main area where you wouldn't be able to practice pronunciation or anything like that out loud.

A more interesting question, I think, is how can those companies whose core resources are not free - websites like LingQ or one of the big boys like RosettaStone - use what they have to help the disadvantaged. Could LingQ find some way to reasonably manage a pro bono tutoring service? The trick would be making sure the students actually deserve pro bono service. Could RosettaStone provide their software to disadvantaged groups? The trick there would be to make sure their donated or discounted software didn't end up back on the market.

Poking around a few websites and a few rudimentary Google searches don't seem to what, if anything, such companies are doing in this regard, so if any one has any examples I'd love to hear about them.