http://blog.fleex.tv/

Fleex premium

Mon, 04 Mar 2013 19:47:00 +0100

As you may have noticed, last Friday night fleex underwent some important changes, mainly moving from a fully free service to a freemium model. While you can still watch videos for free, we took most of our advanced learning tools and put them behing a paywall. In this post I’d like to explain our reasoning and go over some of the changes we’ve made along with the move.

Going freemium

A lot of you were curious about how we actually survived, running a free website with no ads. Truth be told: since fleex started less than a year ago, we haven’t made any money. Zero. Zilch, not a penny. While counter-intuitive, the decision to delay monetization was a very conscious one - looking back, we have every reason to believe it was the right one to make. As it turns out, web consumers have grown very demanding over the years. Creating a service that people actually use has become hard. So hard, in fact, that we figured we needed to put all chances on your side. To do that, we decided we’d let as many people as possible get their hands on what we were coding, and for that to happen we had to start with a free product.

Where did that get us? Well, since fleex’s first beta came out in June 2012 we’ve received more than 800 feedback messages from our users. We’ve paid a lot of attention to those messages, replying to each of them personally. They’ve helped us tremendously in making fleex the great product it is today. Thanks to them, we believe we’ve reached a certain level of maturity, and created a service that’s not only worth using, but also worth paying for.

Fleex uses some of the greatest video productions on the free interwebs, and we’re extremely thankful that it is even possible. The YouTube videos our catalog is based on are available for free outside of fleex, so we saw no reason to stop our users from watching them on fleex. As a free user, you’ll still be able to watch all the videos you want with mixed subtitles, subtitles navigation, notifications and clickable words. This comes free of charge, and we’re hoping that this will help a lot of folks out there to get their English substantially better for free.

For thoses who want to dig deeper, or learn faster, we’ll be offering the more advanced educational tools we’ve put together - the vocab list and exercises, the grammar tree, and the progression dashboard - for a cheap, 9€ / month subscription. Premium users can also expect lots of changes and new features as time goes by - if you’ve been following fleex for a while, you’ve probably noticed that we’re constantly iterating and improving our product. This isn’t going away.

Other changes

Aside from the move to freemium, we’ve made substantial changes to the way our users can authenticate on fleex. We’ve added the possibility to use a simple email / password combination, which should please those wary to use their social accounts. We’ve also merged all accounts created with the same email into a single account. We’ve been very careful about keeping all the data safe: for instance if you’ve added different words in different vocab lists, you can be sure the lists were merged without data loss. Of course, the new merged account can be accessed through any external service initially used to create the duplicates.

Another change we’ve done is we’ve completely revamped the account admin page to let you easily view and modify your personal data. You can now edit your billing info, set up reminders to help you practice regularly, and link more external accounts to your profile in a simple, elegant tabbed interface.

We hope you like it, and can’t wait to find you on fleex!

Grammar comes to fleex

Thu, 14 Feb 2013 12:20:00 +0100

We’ve been pretty quiet for a couple of months. When that happens, you can be pretty darn sure something big is in the works. Step by step, we’re making fleex a more advanced English learning platform, and today I’m happy to announce that we’re bringing grammar to fleex. Like much of the educational features on fleex, grammar is opt-in. If you feel that your grammar doesn’t need us, don’t expect us to come nagging you during your video. However, if you’d like to let fleex help you I’ll explain how we can do that.

The Grammar tree

We’ve added a new ‘Grammar’ menu item to your home page and menu: this will lead you to your personal grammar tree. Your tree is to grammar what the vocab list is to vocabulary: a central place where you define and review the things you’d like to learn. In this case, we’ve worked with a team of linguists to create and sequence a list of all important notions in the English grammar. We cover a large span of concepts, from modals to linking words to various tenses. Every notion is given with its corresponding Common European Framework of Reference for Languages (CEFR) level, so you always know where you are in terms of difficulty.

The first thing you can do with the notions is to simply consult them. Clicking any notion in the tree will bring up a video explanation of the notion to help you understand what it’s all about. As we continue our work on the gramar tree, we’ll textual explanation, examples, etc.

Following notions

The next thing you’ll see are the 2 buttons on each notion. The ‘mark as mastered’ button is pretty straight-forward, and lets you mark a notion as mastered. Mastered notions appear in a green color, so you know your objective: paint that tree in green! The ‘follow’ button deserves more explanations: what it does is it lets you track a notion in your videos. Say you want to work on the present continuous: by tracking this notion, you tell fleex to select occurences of it in the videos you watch and notify you when they come up.

This way, you get to see real-world examples of the notion and can quickly understand how to use them. Like many of the latest research reports confirm it, we belive this is key to learning a new notion - our brain is just that good at assimilation, it would be a shame to stay in the old paradigm of learning (and forgetting) rigid rules. Sheer memory is not enough to be a good speaker.

You know what - that’s enough talking. I’ll just let you get your hands on all this and see for yourself. And if there’s anything you’d like to say, there’s always that ‘feedback button on our site!

Fleexing your videos just got better

Thu, 20 Dec 2012 19:34:00 +0100

Letting you use your own videos has been a bit like Apple’s interests for TV - a project we have a thing for, but more of a hobby, really. Just like Tim Cook publicly changed his mind recently, we’ve had our mindshift. These past weeks, we’ve worked really hard at getting rid of everything standing between you and the ‘fleex your own video’ experience. Here’s what we’ve changed.

Better Compatibility

Until now, the ability to fleex videos had been a priviledge reserved to Windows users only. Not anymore: we’ve extended the feature to Mac users as well, which means for instance that a lot more students (a large portion of whom are MacBook users) will be able to use fleex with their own stuff.

We’ve also made sure that ‘fleex + your videos’ works on all major browsers: if you can watch catalog videos, you should also be able to watch your own material.

An Easier Way to Choose your Videos

Our previous system required you to drop a video file onto the fleex web page, then indicate the path to your video. This was frustrating - having to write down the file path felt redundant - and somewhat complicated - ever tried to get a file path on a mac ?

The solution was to create tiny desktop applications. Who needs a file picker when we’ve got Windows’ Explorer, or the Mac’s Finder? I don’t know about you, but when I want to play a video with VLC I just open the Finder, navigate to my video, right-click it and select ‘Open With’ > ‘VLC’. We had to make it just as easy, and that’s what we did.

Fleex for Mac

Our Mac app isn’t fancy - it’s efficient. It doesn’t even have a graphic user interface - it doesn’t need one. We tried to go minimal on that one, and limit the features to only what’s needed. There are 2 ways you can use to fleex a video:

Right click a file, click ‘open with’ and select ‘fleex’
Go to your ‘Applications’ folder and drag the fleex icon to your dock. Now everytime you want to fleex a video, simply drop it into the fleex box.

And… that’s it. No pain. No headaches. Works out of the box.

Fleex for Windows

Fleex for Windows isn’t news per se: it’s been out for a couple of weeks now. However, it was very much a beta product. It suffered several bugs, one of them even preventing users running a 32-bit version of Windows from installing the app altogether. This has all been fixed now, and things should go down much more smoothly.

Just like our Mac app, the Windows app is limited to the strict minimum: it simply adds a ‘fleex my video’ item to your contextual menu. So whenever you feel like watching a movie, remember: fleex is only 1 right-click away!

(Way) Better Subtitles Matching

I’ll admit it: until now we’d always felt some sort of discomfort with the ‘fleex your own videos’ feature. We knew it wasn’t very reliable. Our stats said it found subtitles less than 50% of the time - clearly not an outstanding performance.

So we took a hard look at what we had done and went bug-hunting. While this is a never-ending chase, I think it’s fair to say we took most of them down. We also improved the algorithms under the hood to get far better matching results. We’ll be monitoring how well we’re doing with this in place, but I can only urge you to go try it out.

And even if you’re not super into the while English-learning thing: it’s still a really practical way to find matching subtitles for your videos. Why not start with that, and let fleex convince you with all the other things it has to offer?

Conlusion

This is a big release for us, and we’re eager to hear what you think of it. Do bombard us with feedback: as always, you’ll get a personal answer and lots of love. Cos’ that’s how we roll.

Introducing fleex for Windows

Thu, 29 Nov 2012 18:28:00 +0100

Disclaimer: this post is for Windows users only. Mac users will have to wait until the beginning of 2013 to be able to fleex their own videos. Hang on in their folks!

Used to be that fleexing a video was rather complicated. Used to be that you had to manually copy and paste the path to your video in order to be able to watch it. Used to be that the whole process was so frustrating we lost a lot of you guys in the process… Friends, that time is over!

If you haven’t already, just head to http://fleex.tv/FleexYourOwn and download fleex’s Windows application. Done? Now restart your favorite browser, right-click any video and click ‘fleex my video’. You’re all set!

Sweet, ain’t it?

Disclaimer

Just like any fresh release, the fleex application is still in beta version. If anything goes wrong with your experience, please make sure to let us know using the ‘feedback’ tab at the top left of the fleex website!

Vocabulary Exercises

Thu, 29 Nov 2012 18:06:00 +0100

Last Friday we released a new version of fleex, and for the first time we introduced vocabulary exercises to the platform.

Three different types of exercises

To start with, we focused on 3 types of exercises. We’ll add more with time, but here’s what we have so far:

‘Fill the blank’ exercises give you an English sentence with a gap in it, and require you to fill that gap. In addition to the context of the sentence, we provide you with definitions of the missing word to help you guess what it is.
‘Find best translation’ exercises give you a word in English and require you to find the best translation for it
‘Match definitions’ exercises require you to look at a list of 8 words and pick the best match for each of the 4 definitions we present you with

Space-based repetition

There’s been a fair bunch of studies on the inner workings of memory. One of the conclusions that is widely accepted is that the optimal way to effectively remember something is to repeatedly be reminded about that thing.

Exactly when to be reminded really depends. Let’s say you’re trying to learn a word: the theory affirms that the first reminder should come shortly after you first encountered the word. Once you’ve got the word firmly implanted in your short-term memory, you can stop thinking about it for while. Wait a couple of days, just enough not to forget it, and then practice again. This time you’re implanting the word in your middle-term memory. Once you’re there, pause again - for a longer period, since you know the word better. Then practice again. You get the gist.

At fleex we’ve implemented space-based repetition at the heart of our exercises. Here’s how it works:

When you add a word to your vocabulary list, it is immediately marked as available for you to practice. If you head to the exercises page, you’ll be able to work on it until you reach the ‘short-term’ level of memorization.
Once you’ve reached the short-term memorization level, the word gets ‘deactivated’ for a while. After you’ve waited long enough, we then reactivate it to let you practice it again. The objective, this time, is to unlock the ‘middle-term’ level.
We repeat the same process for the ‘long-term’ memorization level. Once you’ve reached that final level, we consider the word to be mastered and stop suggesting exercises

This system allows us to teach you words that are here to stay - the old times of huge vocabulary lists transiting in and out of your brain are over!

Disclaimer

I’ll finish with a disclaimer: even though we’re seeing more and more people using fleex, we’re still keen on using the ‘Lean Startup’ methodology as much as possible. This means we delivered on the idea of exercises as fast as possible - as a consequences, you may encounter bugs from time to time. When that happens, please make sure to report it using the corresponding button, at the bottom left corner of the exercises window:

RIP, vocabulary test

Fri, 26 Oct 2012 11:26:00 +0200

At the beginning of this week, we got rid of our “vocabulary test”. The idea behind this feature was to test your knowledge of the English vocabulary. Not that we completely abandoned the idea of assessing your level (see our previous post), but we’d rather take a different approach.Here is why and how.

It’s just too complicated!

Here is a quick reminder of what the vocabulary test looked like:

General feedback was that the test itself was pretty straightforward: if you know the word, click “yes”; if you don’t, click “no”. The problem, though, was in what we did with the results. As it stood, the test brought little value to our users. Statistics were difficult to understand for most users. No clear action was suggested in conjunction with the conclusions we could draw. As a result, we decided to pull the plug and give ourselves more space for a compete overhaul.

Focusing on things that matters

Here are the few criteria that we’ll focus on for the next version:

Objective: probably the only criterion addressed by the previous version. With regular school exercises, you can’t get away from subjective grading: you write an essay, you have an oral exam… Not that it’s bad in itself, but we believe that on fleex, we can provide you with undisputable facts to help you know exactly where you should put your efforts.
Easy understandable: clearly a weak point of the previous vocabulary test. Not that you cannot hide complex algorithms or methods, but they should be under the hood. Very quickly, the user should be able to understand her progress and take actions according to that.
Adapted (or personalized): that may seem obvious but that is probably the most complicated criterion. Why would a user loose her precious time doing exercises on something she already knows very well? The user should continuously be challenged so that she always learns something, but not too much so that she does not feel discouraged.

Would you agree on that? We count on you to challenge us on each of those points!

Work in Progress

Fri, 26 Oct 2012 11:00:00 +0200

Are you doing well?

Fleex’s primary mission is to help you get better at English. We, obvioulsy enough, believe our method works. But most importantly, we want you to be convinced. We want to give you means to measure your progress, and get a sense of how you’re doing. To that effect, with this week’s new release we’ve added a new ‘progression’ panel - in this post I’ll go over its inner workings.

Your progress

Here’s what it looks like - the new page that will help you track your progress on fleex:

We split the panel in 3 parts:

1displays your overall progress in English with fleex. It gives you a quick overview of what you have done so far. And remember this post? It has to be easily understandable and objective.

We show you the time you spent watching English videos on fleex, the number of words you have learnt and the number of words you are currently learning (e.g. words in your vocabulary list). We have several ways in mind to further improve this part, but that will be the topic of another post.

2helps you track your daily activity. Learning a new language requires a regular practice and that is precisely what we insist on here.

Day by day, you can see how many videos you watched, how long they were and how many words you practiced. You can also sign up for daily email reminders. Another thing we plan on doing to enrich this part is letting you set up your own goals.

3shows your journey to the Holy Grail: watching videos in English without subtitles.

Depending on your level, the curve starts at a different level but in general, the curve should go up. Of course videos have very different levels of difficulty, and you may like fiddling about with the subtitles repartition so you will probably notice swings in your curve.

The more videos you watch on fleex though, the more you will understand and the closer you’ll get to watching videos like any English speaker.

Of course, this panel is only at its 1st version and as always, we’re using your feedback to improve fleex. We will enrich this console in the future to be faithful to our goal: make your learning more efficient and more fun.

Notifications

Fri, 05 Oct 2012 01:20:00 +0200

As any disciplined startupers would do in this day and age, the first task we set for ourselves after deciding to create fleex was to build a Minimal Viable Product (MVP). We uploaded a random episode to Amazon S3, coded all sorts of features that we thought would make fleex a killer product, and launched it into the wild.

Of course, it had tons of shortcomings. There was only 1 episode. Ergonomy was poor. Worst of all: it wasn’t scalable. All of the content we used we had written ourselves: we had in-video notifications that gave contextual information - all written manually. We had exercises that trained your comprehension, vocabulary and grammar skills - all written manually. We even validated each user manually…

Naturally, when we began to work on the public beta we quickly realized that a lot had to go. In our minds it was clear that fleex wasn’t going to be an editing company. Sacrificing scalability was too high a price to pay, so instead we decided to focus on other features: subtitles mixing and clickable words.

Today we’re really happy to bring notifications back to fleex. Notifications are a key element to our method. They solve an important problem for our learners, which is that there’s more to a language than grammar, or vocabulary. There are times when you get all the words, get the structure, and yet the whole thing still doesn’t make sense. Notifications help by going beyond individual words to explain groups of words. Expressions, idioms, proverbs, all these locutions that beginners have little chance to figure out on their own. Fleex warns you when they come up, and displays an explanation when you pause the video or hover the notification:

Soon you’ll be able to save the expressions just as you can do words, to practice them later.

What do you think? Is this something you might use? Go ahead and try it yourself!

http://beta.fleex.tv

Subtitles Navigation

Tue, 02 Oct 2012 20:07:00 +0200

With today’s release we’ve substantially improved the way subtitles work. There were several problems with subtitles on fleex, so we decided to focus on 2 of them and tried to solve them to the best of our ability. In the following few lines I’ll explain what changed, and why it matters.

Easy access to easier subtitles

The first problem has to do with subtitles level. While we’ve worked hard to make our mixing algorithm as efficient as possible, it sometimes happens that a line appears in English while you’d rather have had it in your native language. In that case you still have the option to click words and try to understand what you’ve missed, however this is an incomplete solution for at least 2 reasons:

Individual words ignore the global context of the video
If you’re an advanced speaker struggling with an ‘empty’ subtitle, you’re out of luck - there’s nothing you can click for help

We solved this by changing the behavior of the video when hitting the ‘pause’ button: doing so now displays the current subtitle in an easier format, to help you better understand things you might have missed:

If you use mixed subtitles (English + your native language), pausing the video will show the current subtitle in both languages
If you use filtered subtitles (English + empty subtitles), pausing the video will display the current subtitle in English

This way you always have easy access to easier subtitles.

Jump from subtitle to subtitle

The second problem is a classic for anybody that ever tried to replay a specific scene in a movie or episode. You see, the seek bar is so small, and the video so long that jumping to a specific moment is virtually impossible - especially when that moment is only a few seconds before the current position.

The truth of the matter is that the seek bar is probably the worst tool for that specific need. The right unit of measure isn’t seconds, minutes, or percents of the video: it’s lines. When you miss something you want to be able to go back to the last line the dude said, or the one before. That, precisely, is what we let you do now.

When you pause the video, we show arrows on the left and right of the screen: these let you navigate in your video, jumping from subtitle to subtitle. You can also use your keyboard: press space to pause the video, and use the arrow keys to navigate. Note that you can also use the arrow keys while the video is playing: for each press on the left / right arrow key, we’ll go back / forward 10 seconds.

That’s it! We hope you like it - go ahead and tell us what you think!

Status Update

Sat, 15 Sep 2012 13:03:00 +0200

Today we released a new version of fleex - in this post I’ll go over the changes we made and the new features we added.

Manually shift unsynced subtitles

We worked hard to make sure that when you fleex one of your videos, the subtitles we present you with match the exact file you submitted. However, you may still encounter situations were your subtitles aren’t exactly synced, and you get that constant and terribly annoying lag between what you hear and read.

With this new version we give you the possibility to manually correct that lag, much the same way popular players like VLC let you do it. If you need to re-sync subtitles, simply hover the ‘shift subtitles’ button in the player’s toolbar. You’ll be presented with a simple selector that will let you indicate by how much you’d like to shift the subtitles. When you’re done, simply press ‘ok’ and we’ll shift the subtitles accordingly.

Keep a history of all your videos

For the lack of a better solution, when fleexing a video we still require that you manually input the path to your file. However, we now save that path for each video so you can watch it later having to go through the usual process. Just click a video under ‘My Videos’, and pick up right where you left off!

…and more!

More changes include improvements in the way fleex mixed subtitles, the possibility to jump right to the next episode after wat
ching a YouTube TV Show, and a new notification system that informs you of new features, right from the site. Rush to your account to discover it all!

http://beta.fleex.tv

Fleex your videos In a 2010 article posted in the New York...

Mon, 27 Aug 2012 20:25:00 +0200

Fleex your videos

In a 2010 article posted in the New York Times, Nick Bilton made the point that becoming a verb was the ultimate goal for startups. With last week’s release we’re going towards that direction, and we think it’s a pretty big deal. Here’s why.

It’s the content, stupid

At fleex we’ve always thought that learning English shouldn’t be boring. This strong belief is rooted in a simple observation: the world is fed on English content, and people seem to love it. Hollywood movies are going strong. American TV shows are insanely popular. People consume videos by the hour, every day. In English. What’s the need for lessons, when you’ve got such a great opportunity to learn the language while savouring you favorite videos?

This is a cool vision - one people easily get excited about. However, it usually doesn’t take long until we get asked a simple question: where do you get the content? Getting the rights to popular shows seems mission impossible. It is, in fact mission impossible. So how do you do?

The answer is: you don’t. First, signing deals with the majors is no piece of cake. With all their firepower, even giants like Netflix struggle to maintain their links with content providers. But mostly, we feel that this is not our job. Fleex is not a content company. We’re in the business of English-learning, and that’s where we want to put our focus. And so after scratching our heads for a while, we decided to go and implement a way for you to fleex your videos. Any video. Let’s have a look at how it works.

Subtitles, done right

The fleex experience relies heavily on subtitles. Have you ever tried to download subtitles for a video? I don’t know about you - I find it god damn awful. Finding the right match for your video is a nightmare, and many times you’ll end up with subtitles that are out of sync. Of course, you can re-sync them manually. VLC, for instance, will let you do that fairly easily. Often though, that’s not enough: if your subtitles weren’t generated with the right framerate in mind, you’ll find that they keep de-synchronizing.

Our new version starts by fixing that. When you fleex a video (see what I did just there?), we take care of the dirty job and fetch all the subtitles for you. Rather than using the name of your video, we analyze the file and compute an ID that identifies your video in a fairly unique way. We then search open databases using this ID. This way if we find something (which we do most of the time), you can be sure that everything will go smoothly.

All of our tools… with your content

After we’ve found your subtitles, we’ll ask you to give us the folder where your video resides on your disk. Agreed - having to copy and paste the path to a file you’ve just selected and submitted to us seems unnecessary. However, we have yet to find a way to do without this: security-wise, web browsers have no reason to expose the full path of a file selected via the file prompt. The VLC web plugin, though, will need that path to play your video and that’s why we’re asking for it. Note that we’ll remember that folder for you: if you keep your video files in a single folder, you will only need to input its path once.

Once you’ve taken this step, just click next and let the magic happen: the full fleex experience builds on top of your video, letting you work on your English with the stuff you really like (it’s yours, after all). Mixed subtitles, clickable words, vocabulary list. It’s all there.

PC users: it’s on, go try it! To the mac users out there, please be patient. We’re planning on releasing a compatible version soon, so bear with us!

RIP, vocabulary picker

Thu, 02 Aug 2012 10:34:00 +0200

With last week’s new release we got rid of the vocabulary picker. The vocabulary picker was first introduced with a simple idea in mind: for every video watched on fleex, you’ll learn 5 new words. As appealing as the idea was, we realized over the past weeks that it suffered several flaws. In this post I’ll go over the reasons of our decision, and explain how we plan to replace this feature.

Popup frustration

It has grown into an almost Pavlovian reflex: when I see a popup, I just want to close it. Turns out our users do, too. With the way the picker was implemented - an overlaid modal that won’t let you watch your video until you’ve selected 5 words or pressed ‘no thanks’ - this resulted in a lot of frustration.

In our tests, we realized that people like to try videos, and go back and forth until they find something they like. The picker made that process long and painful because it got in the way of our user’s main objective: watching the damn video.

Arbitrary words vs. chosen words

Another problem with the picker was that however intelligent we tried to make it, there was still something very arbitrary about the way it chose suggested words. Here’s a rough description of what the picker would typically do for any given video:

order the video’s words by decreasing frequency
place a cursor somewhere in the resulting list depending on your level
pick words of the list around the cursor

We realized some of the issues this behavior entailed when we released TV shows. TV shows have a lot of very oral words in them. Interjections, onomatopeia - strings of characters that are not words per se but will still be chosen by the picker. The “word” ‘shhhhhh’, for instance, popped up when I watched the first episode of Blue. TV shows also have a lot of proper nouns. When watching the first episode of The Guild I saw words like ‘Tinkerballa’ or ‘Bladezz’ pop into the picker’s window.

This of course degrades the overall quality of the method, and pollutes the learning experience. However, it doesn’t justify getting rid of the picker altogether. We could have added filters, to retain only words that can be found in a dictionary. Or created exclusion lists. Thinking about this though, we still thought something was off with the arbitrariness of the overall experience. With all the diversity of our learners’ profiles, how can we be confident that of the 12 words we push, at least 5 will be of interest to them? And by the way: why 5 words? Some may want to learn more, some less. We felt the picker was getting in the way of our user’s freedom to pick words of their choosing.

Data rules

In the end, all that really matters is hard data. While our experiences with real users gave us a hint as to what to do, our usage metrics was what confirmed it all. When presented with the vocabulary picker, we compared the number of people who clicked ‘no thanks’ with the number of people who chose 5 words and proceeded. The result was rather clear: >25% more people dismissed the picker.

For us, that was a clear signal. We had to take action.

Now what?

We haven’t changed on our vision: we still believe that videos are an awesome way to learn new vocabulary. So how will you be able to do that now that the vocabulary picker is gone?

As said before, there is something very arbitrary in the idea of pushing a finite list of words and expecting a user to choose 5. Again: how do you select the list? Not everybody has the same needs, or interests. People with the same overall level might have vocabulary deficiencies in very different areas, so we thought it was important to give more freedom to the user in choosing the words they want to learn.

We also felt we needed to preserve the general notion of pushing vocabulary words to our users: the idea that learning with fleex should be as effortless as possible is key to us, so words had to continue to come to our users - not the other way around.

The solution we came up with uses what we think is the most natural way to push new words to our users: the subtitles themselves. It works in a simple fashion: whenever you meet a word you don’t know, click it. This doesn’t differ from what you already know (clickable words) - what changes is the new ‘add to my list’ button. After reading about the word you clicked, if you feel you’d like to learn that word just click the orange button and we’ll add the word to your list. Once a word is in your list, we can remind it to you when it appears in other videos. This way you practice it without even realizing it.

We think it’s an elegant solution: it doesn’t interfere with the viewing experience, and integrates well with an existing feature. Moreover, it leaves it entirely up to you to chose the words you want to learn.

What do you think?

How good is your English?

Thu, 12 Jul 2012 18:49:00 +0200

Since our last release you have been able to use fleex to improve your English vocabulary. While many of you may already be convinced that videos are a great tool to improve your language skills, measuring exactly how efficient it is remains hard. After a lot of research, we’re proud to announce fleex’s brand new level estimator. In this post I’ll walk you through the nuts and bolts of it.

Testing your vocabulary

First of all, you may wonder why we’re testing your vocabulary. What about grammar? An honest answer is that we’re starting with the easier bits - as you will find out though, ‘easier’ doesn’t mean ‘easy’. As often the case, the devil is in the details.

A naive (albeit correct) way to test your vocabulary would have you go through every English word there is and tell us whether you know it or not. Unfortunately there are about 45,000 used words in the English language, and there’s nothing very exciting about reading every entry in a dictionary to tick the ones you know.

So we had to find an alternative - that’s where it got complicated. Below are a few questions that we found particularly tricky:

What is a word? Let’s say you know the verb “like”. There is a very good chance that you also know the words “likes” and “liked”, as they are just derived forms of “like”. So a first thought would be to count those words as only one word. But then take “could” (the preterit form of the verb “can”). While some of you may know this irregular form, some may not and that goes for all irregular verbs: better to count its various forms as different words. This is a tough problem, with no perfect solution. For this first shot, we went for simplicity and decided to count “like” and “likes” as two different words.
How can we tell that you know a word? When asked if they know the meaning of “keep”, a lot of you will tick the “yes” box because duh, everyone knows that “keep” means “garder”. And you will be right. However, a keep can also be a place where a princess is detained waiting for her valiant knight to save her (“un donjon”). Less people will know that because that meaning of “keep” is less used nowadays. Still, you may see it from time to time and you won’t understand that sentence. Here again, we thought that problem wouldn’t arise all to often and decided to start by ignoring it.
Is knowing the words always enough to understand a sentence? You know the words “rain”, “cat” and “dog”: surely, you must be able to understand a sentence composed of those words. That’s ignoring expressions, like “raining cats and dogs” (il pleut des cordes). Meeting an expression for the first time, it is unlikely you’ll understand it from its words only. This issue will be treated later with notifications (little bubbles during the videos), as it is clearly something which goes beyond vocabulary.

With all these caveats in mind, I’ll move on to the actual solution we implemented and explain how we went with with.

How does it work?

If you log in to your fleex account and head to this page, you’ll land on a new tab called “Mon niveau”.

The tab is split in two parts:

[1] is where we test your vocabulary
[2] is where you can see your score

The test part [1] is pretty straightforward. We show you a word in 3. If you know it (no cheating!) simply click on 4; if you don’t, click on 5.

The score [2] a bit more complex:

[6] shows you an estimation of the number of words you know. Altogether, fleex’s videos contain a limited set of words (around 35,000 as this article was written); the figure in [7] tells you which percentage of all the words on fleex you master.
The gauge in [8] reflects this previous figure: the fuller it is, the better your vocabulary skills are
[9] gives you your chance of knowing any word randomly picked in one of fleex’s videos. It’s not the same figure as [7] as words have different frequencies, and we’re making you work on the most frequent words first.
[10] and [11] are the uncertainties around figures [7] and [9] respectively. As you start, those two circles are red because the uncertainty is high. The more tests you take, the greener the circles. Note that you can always get the exact uncertainty figures by hovering the circles.

Behind the scenes

“Curiosity killed the cat”, but you’re not a cat. For those interested I’ll write a little bit about what’s happening under the hood.

Our test is based on a simple observation: learning a foreign language, you memorize the most frequent words first. There are several very good papers out there covering this very subject - to those wanting to know more about vocabulary acquistion, I can only recommend Measuring Second Language Vocabulary Acquisition (book available here). Below is a graph from the book that shows the vocabulary profile of a typical learner:

As shown by the graph, the typical learner will know 85% of the 1,000 most frequent English words, a much lower 65% of the next 1,000 words, and an even lower 45% of the following 1,000 words, etc.

These ‘bands’ of frequent words are heavily correlated:

the 3 most frequent words of the English language are “the”, “be” and “of”. If you know them, then we can deduce with a good approximation that you also know the word “and” (the 4th word the most frequent) because they are in the same frequency band.
this also works the other way around: if you don’t know the words “sycophancy”, “ecliptic” and “spool” (in the last frequency band), it is very likely that you won’t know the word “daunt” either because they are all infrequent words.

Based on that observation, we split the words in different groups according to their frequency. We order the words from the most frequent to the less frequent. We put the first 1,000 words in the first group, the next 1,000 words in the second group (from 1,001 to 2,000) etc… Then we assess how many words you know in each group by presenting you a few words in each of those groups.

Let’s do an example: say we start with the 1st group, and you mention that you know the words “the”, “be” and “of”. You score 3/3 (you know all 3 words we presented) so we deduce that you know 100% of the first 1,000 words. We’ll say that you do just as well with the second group.

Moving on to the 3rd group (the 2,001 to 3,000 most frequent words), you indicate that you know the words “temporary” and “impressive”, but not the word “junk”: we conclude that you know 66% of those 1,000 words.

Let’s sum it all up: you know 100% of the 1,000 words in the 1st group, 100% of the 1,000 words in the 2nd group and 66% of the 1,000 words in the 3rd group (666 words). In total, you know 2,666 words!

At this stage, of course, this is still a pretty rough estimation. As you answer new questions though, the estimation gets ever more precise; by our measurements, testing 10 words in each of the 12 frequency groups gets you to a very acceptable level of uncertainty.

A means to an end

To conclude, here are 3 things we think are key to this new level estimator:

Frequency matters! The number of words you know isn’t the only indicator - the frequency they cover is important too. It’s especially important when learning new words: learning frequent words first makes you progress faster with the same effort.
The test is 100% integrated with the rest of the website. So if you already used fleex and added a few words to your vocabulary list, you will not see them in your test. Also, as you refine your score by taking the vocabulary tests, fleex will remember what words you know and what words you don’t know, so we never present you with words you already know.
Getting a better estimation of the words you know / don’t know will help us tremendously in improving fleex’s educational experience. With those estimations we will be able to offer new features, for instance estimating the difficulty of a given video for you. Wouldn’t it be nice if fleex could tell you: “you know 98% of the vocabulary used in this video, you should go for it”?

Again, we’re always pleased to get your feedback on these new features we release, so do press that ‘feedback’ button on our site. And stay tuned: there’s a lot more to come, and we’re going to hit you with it in the coming days. So read our blog, or follow us on twitter, or meet us on our facebook page.

Introducing vocabulary training

Wed, 27 Jun 2012 17:10:00 +0200

“5 words a day keep the teacher away”: tonight we’ll be adding vocabulary training to fleex. We’ve tried hard to stick to our vision of making great learning tools that preserve the pleasure of your video experience. We looked deep into some of the most recent research in vocabulary learning to make the experience as efficient as possible, while always keeping you, our user, in mind. In this post I’ll show you around the new features and tell you a bit about the reasoning that went behind them.

Keeping it simple

Behind vocabulary training is a simple idea: for every video you watch, you’ll learn 5 new words. You get to choose the words you want to learn, from a list of words adapted to your level. We keep track of the words you choose, so you can go back to them at any time.

We’ve worked hard to keep it simple: it should really be self-explanatory. In case it isn’t, here’s a few words on how it all comes together:

Each time you watch a video, you’ll be presented with a list of words and asked to pick 5.
With each word you can do 2 things:
- 1 Clicking the word adds it to your list of 5 words you want to learn
- 2 Clicking the little cross next to the word lets us know that you already know that word. When that happens, we memorize your choice so we don’t present you with this word in the future.
3 If you don’t like any of the words we suggested, you can ask for more by clicking the corresponding button
4 Words in your select list have available information on them so you can start learning: translation, definitions, pronunciation… We show you this information right after you select a word, but you can also display it at any time by clicking any word in the list.
5 At the top-right corner of the popup you will see 5 dots - they show how many words you still need to pick. If you’ve already picked 3 words, 3 of the 5 dots will be orange while the orange 2 will remain gray.
6 When you’re done choosing the 5 words, a green tick will inform you that you can move on to the video. Simply click the ‘next’ button and you’re free to go.
7 Vocabulary training is 100% optional: click the dismiss button at the bottom of the popup to continue to the video.

Why it works

Before getting started with the code, we did some research to determine how best to go about it. There is extensive literature on vocabulary learning techniques, and we tried to build some of the results we found into the tool.

Adding intentional learning into the mix

A first distinction we looked into is incidental vs. intentional learning. Recent studies produced compelling evidence supporting a mixed approach, combining incidental and intentional learning. In her 1994 PhD dissertation Dr. Cheryl Zimmerman, a professor at UCSF, found that 3 hours a week of explicit vocabulary instruction plus some self-selected reading were more effective than reading alone. In a book published in 1997, Prs. Paribakht and Wesche also found that reading plus explicit instruction led to superior gains over a period of three months.

This was for us a strong sign that combining incidental learning - encountering an unknown word, pausing the video and clicking it to get its definition - with intentional learning - explicitely choosing 5 words to learn and being confronted with them - would be a winning strategy for our users.

Using dictionaries as a tool

Another commonly used distinction opposes contextual guessing to the use of a dictionary. Again, a mixed approach seems to win the votes of researchers. In an article published in 1994 Dr. Knight, a professor in Psychiatry at MUSC, discovered that while incidental vocabulary learning through contextual guessing did take place, those who used a dictionary as well as guessed through context not only learned more words immediately after reading but also remembered more after two weeks.

We built these findings into the tool by combining both approaches:

displaying subtitles in English creates room for contextual guessing,
showing dictionary info during the 5 words selection process + when a word is clicked in the subtitles gives you the occasion to use a dictionary and enhance your learning experience

A tough decision we had to make was to decide the language in which to display definitions. Some recent research showed us the way, pointing to the use of a “bilingualised” dictionary as optimal. Bilingualised dictionaries provide definitions in the target language (e.g. English) together with translations in the learner’s native language. In an article published in 1997, Drs. Laufer and Hadar assessed the effectiveness of monolingual, bilingual, and bilingualised dictionaries with 123 English learners in Israel. They found that irrespective of the learners’ proficiency level, the bilingualised version was either significantly better than, or as good as, the other two types in both comprehension and production tasks.

Measurable progress

What may seem an obvious conclusion is also consistently backed by research: getting a sense of progress plays a key role in motivation. In a recent study highlighted by HBR’s ‘Breakthrough Ideas of 2010’ report, researchers asked knowledge workers to email them daily diaries describing the events of the day, and their overall feeling for that day. A close analysis of the nearly 12,000 diary entries revealed that on 76% of the ‘best days’ reference was made to progress being made. The second most cited reason, instrumental support, was only cited 43% of the time. Based on that conclusion, we set it as one of our main objectives to make progress measurable.

The new vocabulary training feature is a first step in that direction. There’s a simple key: with it you’re not just learning words - you’re growing a list. The bigger the list, the better you’re getting. A simple way to get a sense of your progress is also a powerful motivator.

The future will bring more: a number of words is the beginning of a metric, but we want to go further. Some words are more frequent than others - learning them can help you understand way more things than others will. Knowing the meaning of ‘house’, for instance, is a much bigger step on the learning curve than discovering what ‘sycophancy’ means. As a word, ‘house’ unlocks way more conversations than ‘sycophancy’ will ever do. In future releases, we’ll introduce other measures of progress that factor this in.

Credits

A lot of the references found in this article are drawn from a paper written by Peter Yongqi Gu and published in 2003 in the “Teaching English as a Second or Foreign Language” Electronic Journal. Peter’s paper reviews empirical research on vocabulary learning strategies in a second/foreign language. You can read it here.

Fleex’s story Today we met with Moran Bar. Moran is the CEO of...

Sun, 17 Jun 2012 19:00:00 +0200

Fleex’s story

Today we met with Moran Bar. Moran is the CEO of Venturegeeks, a start-up accelerator program in Israel. She has been an investor in seed stage startups for a few years now, and she gave us valuable advice I think on what investors look for in a startup.

During our conversation, Moran touched upon the importance of stories. Stories can ground big conceptual ideas in to the reality of our everyday life, and that’s why founders with a story usually make a much more compelling case for their startup. Thinking about what our story was, I thought I would write a bit on our motivations to build fleex.

Genesis

A cliché is a cliché for a reason: as unoriginal as it sounds, I was standing in my shower when the idea of fleex hit me. The previous night I’d had a conversation with friends about how English was invading our lives, and how little we could do about it. France has long fought for the preservation of its language and everything gravitating around it - the precious “francophonie”. But the French zealotry has had little impact on the advance of English as an international language. Countering globalization is an unrealistic objective - instead of swimming against the tide, France should embrace the trend. “If you can’t beat them, join them”, says the word. We all agreed.

The consensus on how exactly to achieve that was less clear. We (or rather our parents) had tried a bunch of things, and sharing our stories was a joyous experience. There were the English lessons you get from school, and the endless and oh-so-boring debates on climate change or sex parity. Or the ridiculously overpriced 1-week, intensive programs whose benefits vanish as fast as they come. Or the ‘fully immersive’ stay with a family that receives an average 35 students every summer and have long lost any interest in talking to them. Or even the pricey summer camp, where you end up gabbling pseudo-English for a month with your fellow foreigners, none of them English or American (that is, if you have the courage or will to cut yourself from your ‘national’ group, which most teenagers don’t).

This, of course, is a bit of a caricature. After nearly 10 years, I still see the family I was sent to every year. They’re great and I love them (hello there Underhills - I’m talking about you!). Still, I’m one of the lucky few. For the most part, today’s methods are inefficient and too often expensive. It shows: according to an article published in 2009 by French newspaper “Le Monde”, France’s scores at TOEFL, an English test, place the country at the 69th rank out of 109 countries. Looking at Europe only, France comes 25th of 43 nations. Where does that leave us? I didn’t have to look too far away to find the beginning of an answer.

And then…

I was introduced to American TV shows by my younger sister. One of her friends had lent her the first season of ‘Lost’, and she insisted that I had to watch it. A couple days later I was addicted, spending entire nights watching episode after episode. It didn’t take long before I started taking on other shows. The quality of American productions is unmatched. In less than 2 decades, TV shows have become super productions with budgets in the millions of dollars. Networks finance dozens of new pilots every year, keeping only the most successful. I got so addicted to TV shows I had to have them as soon as they got out. The only solution at the time was to download them, but that meant no subtitles - waiting the couple of days it took fansubbers to make them was not an option.

I learnt so much with TV shows, you’d be amazed. Tons of words. Expressions. Idioms. Every week came another episode, reminding me of what I might have forgotten and introducing new things. The best part is, I wasn’t even noticing how much progress I was making - I was having so much fun. Not all learning has to be painful. Slowly, my ear was getting accustomed to the accent, and my oral skills were increasing dramatically. Bits of English were flowing in, attached to memorable images and emotions that made them really sticky. Weekly episodes meant regular practice. Cherry on the cake, I was learning tons of things about the American culture.

All this came back to me as I was slowly waking up in my shower. There was an “eureka” moment - if the method works, how come so many people ignore it? There was something here, and I wanted to be part of it.

Reality check

Searching the subject and talking to people, we soon got some really positive results. A professor at a respected Parisian school confirmed to us that TV-show lovers speak with a better accent than other students. We found research backing the efficiency of videos and subtitles as learning material, and studies like this one correlating the excellent level of English in Northern countries to subtitled original version broadcasts (as opposed to dubbed television). We even learnt of students passing deals with their parents to be allowed to watch tv, provided they used English subtitles. This proved not only that the method worked, but also that our intuition was hitting a pre-existing trend.

Of course, there were gray areas. Getting good content is tricky. Building a solid method requires knowledge on learning dynamics. But we concluded we had something - and that, my friends, is how it all began.

A few words on Le Camping In the beginning of June Fleex was...

Fri, 15 Jun 2012 15:23:00 +0200

A few words on Le Camping

In the beginning of June Fleex was selected for season 3 of Le Camping, a French startup accelerator based in Paris. From the early days of coding at home to the shiny open space we now share with 11 other startups, it’s been quite a change. A couple of weeks into the program, I’d like to share some of our first thoughts and impressions.

Context

The debate over whether Europe can ever catch up with the American startup ecosystem is a classic one. The last few years have seen a lot of changes in this area, with notable initiatives like Seedcamp, Startup Bootcamp or Springboard. In a discussion paper published in June 2011, NESTA, a charity promoting innovation, featured 9 accelerator programs in Europe. France hosts none of them - a void Le Camping wants to fill.

If any conclusion can be drawn from the selection process alone, there is a lot to say on how much Le Camping differs from other French initiatives towards entrepreneurship. In 3 rounds spread over a short 8 weeks, 200 startups are screened to retain only 12. The recruiting team uses modern technology to communicate with applicants along the way: the application form is submitted on f6s, a slick portal for entrepreneurs, and interviews are conducted on Skype or Google Hangouts. No trace of paper can be found. Compared to other contests in which we’ve participated, the experience was considerably more pleasant, and the process more efficient.

Then comes the question of selection criteria. France seems to have an obsession for labs, research and hard technology. Starting a biotech firm makes you much more likely to receive public funds than building a website, however innovative. The French administration also has a thing for sound business plans and safe revenue streams. While these criteria work well with industrial firms, the same can’t be said of internet startups where usage innovation prevails and many projects start with an unproven business model. This puts web companies in an uncomfortable situation, with teams bullshiting their way through applications and interviews with people who, well, just don’t get it. Le Camping takes the right approach, borrowing much from successful American programs like Y Combinator or TechStars. In the context of a seed stage internet startup, their questions focus on the right things: who is the team made of? Have they built anything before? Do they have what it takes to reinvent themselves and pivot if the market proves their intuitions wrong? We really had an ‘Ah…’ moment there. It was the right spirit, and much of it reminded us of our days at Stanford.

First impressions

When the program began, we had just released fleex’s public beta. The first weeks being strongly focused on need finding, it was the occasion for us to stand back a little, keep our hands off our keyboards and think about what problem exactly we were solving. It encouraged us to create what we’re trying to develop today: an open environment where users are encouraged to give their feedback and contribute to the creation of the ideal product.

To be fair, all this wasn’t completely new to us. We were already familiar with the ‘Lean startup’ concepts, and much of what we’ve been hearing in mentor sessions so far we’d heard or read before on places like Y Combinator’s Hacker News. Still - it never hurts to hear it again. Being your own boss, the temptation is strong to ‘go your own way’. As much as I love Fleetwood Mac, and unless you’re an absolute genius (which most of us aren’t), this isn’t the safest route to success.

Much of the knowledge we receive is given through speeches, but there is practice involved as well. Last week, we were asked to ‘get out of the building’ and gather information from real people. In a ‘Design Thinking’ workshop, we were required to impersonate an imaginary user into a ‘persona’ and draw her typical journey. Then we reflected on the perfect user experience for that persona. Such practices are not always very helpful - it really depends on the product you’re building. Whale Street, a Camping startup and our neighbours in the office, is devising a way to predict stock prices with Twitter. Their product is strongly B2B. Gaining insight into a fund manager’s daily life matters far less than finding the formula that will yield accurate forecasts. For B2C products, however, the experience can be eye-opening. Our conclusion on fleex was that people would use our product either as a daily routine or as a time-filler. Our objective was therefore to focus strongly on reducing the friction and simplifying the user experience. The faster and easiest it is to sign in and get to a video, the more people would use the product.

Teaching isn’t all. Le Camping is also a great contact enabler, granting us access to a thriving community of entrepreneurs and investors. Each team is assigned a lead mentor, picked from a list of about 25 - and you can always meet other mentors if their expertise meets your needs. Most mentors have high profiles: Andrew Lacy was co-founder and COO of Tapulous, a gaming company acquired in 2010 by Disney, and is a lead mentor this season. We also get to meet important figures of the tech scene: at a ‘Google Off Site’ event organized last week, an international delegation led by no less than Dan Teodosiu, Engineering Director of Google’s offices in Paris, was here to listen to our pitches and advise us. As a company relying strongly on YouTube for video, this was a valuable introduction.

Conclusion

It is probably too early to tell whether or not, as our friends at Poutsch (a Camping startup) like to ask, Paris will be the next Tech hub. Le Camping seems off to a great start - only time, and the number of candidates that eventually grow into global, successful businesses, will tell.

Fleex is on Tumblr! We’re really excited to be launching fleex’s...

Thu, 14 Jun 2012 16:39:00 +0200

Fleex is on Tumblr!

We’re really excited to be launching fleex’s blog on Tumblr today. Being able to communicate on the product and tell you guys a little bit more on what we have in mind is certainly an objective for us.

What will you find on this blog? Many things. Release notes, how-tos, deep thoughts on not-always-so-deep subjects… If you want to engage with us, tell us what you think, what you want, the things you expect from fleex, this is also a place for you to get heard through comments - we’re usually pretty reactive, and we like to keep it personal. We’re humans after all ;-)

After Twitter and Facebook, Tumblr is yet another step in a effort to be present wherever you might want to find us. We’ll try to post regular updates - bookmark us today and we’ll keep you posted!