Building a Better Language App

Over the past three years, I’ve taught English in three countries while learning three languages. These experiences have been wonderful opportunities to meditate on and tinker with existing methods.

With each country, I faced a different culture, classroom setting, and set of needs. My teaching style had to morph to match the circumstances. Sticking to one method would only risk alienating students. In Chongqing, I oversaw twenty-two conversational English classes at a junior high school, each of which had 60 students or more. I had complete control over the curriculum, but was required to create Keynote presentations (and have their content vetted by the head teacher). Most of my day was spent trying to get through it. Prague was an experimental period. I tram-hopped to various companies to teach business English to small classes of 20-60 somethings. Since most of my students despised textbooks and had specific goals in mind, I had to create most of my lessons from scratch. In Taipei, I had to follow stricter guidelines. For every intensive, general, and test-prep class I taught to adults and high school students, I was required to use a specific textbook at a certain pace and attend regular workshops.

Where training failed, language study stepped in and made adapting easier. After failing to learn Czech in a classroom setting during my initial months in Prague, I committed to pursuing my interest in another language—Japanese—on my own. Once I refined my approach and techniques, I easily moved on to Mandarin Chinese and Russian, and even returned to Spanish.

By teaching in a classroom while learning on my own, I discovered gaps in traditional methods, possible alternatives, and new materials. And my teaching evolved further.

Leaving education for iOS development isn’t leaving at all. It’s simply a continuation of a three-year experiment, with the added challenge of linking together everything I’ve learned to create awesome language apps.

Unlike most teachers, I believe mobile can do language education better than teachers in a classroom. It has the potential for greater consistency, personalization, interactivity, and accurate feedback.

Yet most language apps fall short. Instead of offering new approaches to learning, they merely replicate existing ones. Once you look past the medium—the novelty of learning languages through apps—it’s clear that most companies are simply digitizing age-old techniques. Most use translation and matching activities, listen & repeat pronunciation drills, dialogue gap-fill exercises, topic-focused vocabulary, and flashcard review. These techniques are old hat for teachers. The only difference is that they occur in shorter timeframes and include a point or badge system in lieu of tests. Sometimes they appear as clever games, as in the case of Cat Spanish. This has done a lot for access and motivation, but not so much for education. The silent assumption in apps like Duolingo, Babbel, and Memrise is that (a) the main problem in education is access and (b) creating apps that replicate existing methods will replace, or at least diminish the role of, teachers and therefore solve that problem. Hooray.

Unfortunately, the only part of this assumption that’s true is (a). Access is a huge problem. But, so is crap teaching. Replicating ineffective techniques on mobile doesn’t magically make them effective, even if doing so is technically impressive.

The best apps will investigate what mobile can do for language education that teachers can’t, won’t, or haven’t thought of doing, and develop accordingly.  Finding the answer is surprising simple. You just have to figure out which skills teachers struggle to teach, fail to teach, or don’t teach well consistently (due to fluctuations in energy and other human problems).

Since I’ve been a teacher and spent loads of time observing others, I can identify these problem spots quite easily. Without a doubt, the hardest areas to teach are pronunciation, listening, reading, and culture.  The first three are difficult because they require strong attention to detail, anticipation of students’ needs, adaptation to different learning styles in the classroom, and varied feedback. The last is either totally ignored by teachers who assume that culture comes later, or introduced as an aside without any practice. Plus, identifying cultural nuances is difficult; teaching them in a structured way is even harder.

Why does this matter? Mastering these areas is essential to becoming fluent. Good pronunciation creates confidence (and leads to positive reception from locals). It’s also tied to listening, which is essential for conversation skills. Reading is one of the best ways to expand vocabulary and learn about the culture. With some languages, namely Spanish and Russian, literature is a cultural touchstone. And understanding culture is essential to communicating, especially at a high-level.

So, if teachers sometimes fail in these areas, how does mobile do? Not much better, and in most cases, worse.Speech recognition doesn’t give users feedback beyond whether it’s correct. As such, it’s next to useless. If users don’t know why they’re mispronouncing a word, they won’t improve. Mondly has tried to go beyond basics by actually dictating what you say, exactly as you say it. But it doesn’t identify the specific sounds users need to work on, let alone offer drills based on those needs, like a good teacher would. Babbel is great at teaching the details of pronunciation—each individual sound—but falls short when it comes to teaching natural, conversational speech—how those sounds change when smashed together in rapid conversation. Some services, namely Mango Languages, Babbel, and Innovative Language 101, do well with introducing culture, but not in offering an opportunity to apply it. And very few offer a structured approach to reading. (There’s Bliu Bliu, but it’s not user friendly.)

I could go on. But these are just details. Let’s look at some of the larger problems.

The vast majority of language apps lack the following:
  • Question-centered learning
  • Intimacy
  • A user-tailored curriculum
  • Multiple layers of translation
  • Social & nonverbal feedback
  • Surprise
  • Alternative methods for review (read: not flashcards)

Addressing these gaps isn’t as difficult as one might imagine. Let’s look at why they’re important.

Question-Centered Learning

Language apps are typically designed around vocabulary, with grammar as a secondary focus. In this sense, they’re like a grammar-light version of textbooks. The decks you see on Memrise and Anki, and the courses you find on Duolingo and its language-specific imitations are organized around topics (vocabulary), leaving users to learn grammar through pattern recognition.

From a business perspective, this makes sense: students turn to apps because they’re either frustrated with traditional courses and textbook material, or they want extra practice and support. Designing courses that are fun and grammar-light keeps both groups happy: the former gets something different, and the latter avoids the dull grammar drills of the classroom.

In terms of learning, though, it’s not effective, so it’s a short-term business strategy at best. Users will drop as the novelty wears off and the repetition starts to feel just as tiring as old methods.

But the flaws of this approach run deeper. People learn languages to have conversations, which are centered around questions. Most language apps do nothing to get me closer to having a conversation, especially in the first few lessons, which typically focus on identifying people and objects. Imagine walking up to someone on the street or at a party, and saying, “I am a woman”, as Duolingo teaches you to do in its first lesson. That person would either think you’re insane, into stating the obvious (assuming you’re a woman), or trying to undermine gender distinctions (if you’re a man and they’re intellectually sophisticated). If they have a sense of humor and/or an interest in 1970s-style feminism, they might quip, “hear me roar”. Regardless, they would probably not become your friend.

Perhaps I’m being unfair. How can I dismiss an entire language learning method based on the first few lessons, especially when most apps have sections for questions? Why are questions so important anyways?

If user data is any indication, I’m being more than fair. Considering I spend up to a week on a language app before rejecting it, I’m even being generous. Most users, overwhelmed by life and a constant deluge of information, accept or reject apps within seconds. So, being useful from the first lesson is essential.

Since questions are required for conversations, and conversations are the cornerstone of fluency, language apps should be designed around them. Users should be equipped to have conversations from the get-go. The vocabulary and grammar of the language will follow.

You can learn a lot about a culture from its most common questions. Americans’ fixation on career and status is apparent by the fact most people will ask ‘What do you do?’ within the first three questions. A question like this would be considered tasteless in other cultures. When I was in mainland China, the dominant question was ‘Are you married?’, which would be considered offensive or irrelevant to many westerners. In the Czech Republic, I was often asked about my education. My students became warmer when they realized I was teaching English as a choice, not due to lack of education or job opportunities. In Taiwan, the most common question is ‘Have you eaten?’ or ‘What did you eat?’. Food is a source of pride here; it’s an essential part of the culture.

To its credit, LanguagePod101’s Top 25 Questions series is centered around questions. But it’s too formulaic and limited in scope to empower users to adapt to multiple situations. Plus, it offers no cultural context. Users have no idea when they should ask these questions or why, which means they’ll default to the conversation rules and patterns of their own culture, which could be disastrous. Many of the questions, such as “How old are you?”, would be considered odd or inappropriate in the culture, and are consequently useless. The question “How are you?”, for example, might be a common icebreaker in the U.S., but it’s rare in other cultures.

The Case For Curiosity

My fixation on questions goes deeper than my interest in teaching effectively. The questions aren’t the point; it’s about keeping students curious.

Why does curiosity matter?

For one, you learn better when you’re curious. Second, by asking questions, you train your brain to default to investigating rather than judging. When you live abroad or work with other cultures, there’s always the risk of becoming more narrow-minded as a result of one bad experience. Study abroad programs and Tumblr users might like to imagine that travel always widens your worldview, but this is only true given a pre-existing curious disposition (and having your basic needs met). What separates people who overcome culture shock from those who don’t is curiosity. Curious people can use questions to step back from snap judgments and reevaluate the situation.

A Possible Solution

So what would a language app centered around questions look like?

I’ll offer a concrete example.

If I were to redesign Duolingo or any language app out there, I’d turn each level into a question and arrange all grammar and vocabulary accordingly. Users would learn questions and possible responses, and only then would they move on to additional vocabulary and grammar (all of which would be related to the question). Then I’d put the questions in the order they typically occur in a conversation. Each lesson would be no more than five minutes and have options for lexis and grammar extension (for motivated users with more time on their hands).

The questions and the order of them would reflect the quirks and interests of the culture. You would have the option of choosing different scenarios as well—party, travel, work, dating—since questions vary based on situation. Practice and review sections would be conversations. In the first level, users would simply choose the missing word or response; in the second, they would write it; in the third, they would say it; in the fourth, they would put the conversation in order. There would be no flashcard or review feature. Instead, material would build on previous lessons.

With this approach, users would immediately be prepared to have conversations—not at the end of the course, but from the first lesson. And that’s more empowering than points.

Intimacy (And The Problem with Gamification)

This brings us to the larger problem of the gamification model. I’ll admit that Duolingo and Memrise’s point systems and leaderboards make the learning process addicting. That’s why I recommend using them as a warmup for longer study sessions.

But then it starts to feel like school, and I stop using them.

In fact, I’ve never used a language app that resembles the experience of being in a new country. What would it be like to use an app that did? To feel like you’ve just stepped off the plane?

Imagine that you’ve just arrived in a big city. Let’s imagine that it’s Tokyo because, let’s face it, the try! Swift conference in March is gonna be awesome. You have no experience with the language or culture of the country. Thanks to the sleep-deprivation from the flight and being in an unfamiliar location, your hearing is sharper than usual.

What do you hear? An unfamiliar language, but not as you would in an app or audio phrasebook: words surface as mumbles, laughs, whispers, and clipped phrases. These human voices are occasionally punctured by courteous warnings from escalators and crosswalks. You would feel like you’re in a 1980s Sci-Fi movie, if it weren’t for the more familiar sounds: the click-clack of high-heels, the opening of umbrellas, rain, ringtones.  The video Tokyo City Symphony captures this experience perfectly.

Every city has a distinct sound, energy, and color template. (That’s why so many people on SoundCloud, including yours truly, record ambient noises during their travels.) Tapping into that is essential for any developer who is looking to create, not just a learning area, but a cultural headspace. Every time your user opens your language app, their orientation to their environment should change. They are no longer in their home country; they’re in another. Immersion by design and through technology.

Taipei, Taiwan
Chongqing, China
Prague, Czech Republic.


There’s a reason why Monument Valley and Kami don’t need to offer points or badges. I’m sure I’m not alone in finding them more relaxing than meditation. My entire mood evens out when I play them. They’re entry-points into another world.

The difficulty with language apps, of course, is balancing users’ fantasies of a culture—built from popular films, literature, music, and whatnot—with the sometimes ugly reality of it. A little romanticizing is necessary, but not so much that it promotes stereotypes. This gets into knowing the assumptions you’re users are going to have, and subtly working to change them. A challenge, for sure.

Where’s the personalization?

Given all of the talk about personalization in tech, there isn’t a lot of it in language apps. If you’re a 26-year-old Mongolian-American female who speaks Dutch and likes hockey, you should learn how to say that first. Mobile is already equipped to adapt to user input. iOS developers should spend time thinking about how to arrange content with this in mind.

Layers upon layers of translation

Language apps need varied forms of translation. Some have already recognized the need for literal translations, which are helpful for understanding grammar mechanics. But, we could go deeper and remove English entirely. What about using emoji? What icons or graphics could make the meaning and context clear?

 Social & nonverbal feedback

When you’re having a conversation in another language, how do you know when you’ve made a mistake? The other person’s facial expression. A look of confusion or disgust. If you’re blind, you might know based on the length of pause. What do language apps use? The same stuff teachers use: check marks, points, and colors. If apps are going to successfully replicate the experience of having a conversation in another language, and they should, the expressions on the faces of icons needs to change during chat practice. We’ll examine how to do this in the next article.


During the first two weeks of using Babbel, Duolingo, ChineseSkill, and CatSpanish, I was absolutely addicted. But then the repetitive structure of the lessons started to become boring, and I quit. Apps, in short, need more variation and ways of jolting users to sustain interest.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s