inContext at a Glance…

inContext is a novel AI dictionary that creates unique, context-informed definitions specifically tailored to your language level. Though it’s principally aimed at intermediate language learners, it can be useful for anyone who wants to grow their vocabulary in a target language. In this post, I will explain what the tool is and why we made it.

What is inContext?

inContext is a browser extension. Here’s how it works:

• While browsing, you encounter a word you don’t know. You select it.
• An AI language model creates a unique definition for this word in the target language based on the surrounding context and your language level.
• You quickly read this definition and continue browsing with minimal interruption.

inContext solves three problems posed by traditional dictionaries for language learners:

1. Standard dictionaries often lack context, offering multiple possible definitions and leaving users to determine the most applicable one.
2. Definitions are usually written for native speakers, so the vocabulary and terminology is very advanced and can be overwhelming for learners.
3. Looking up words in a traditional dictionary is time consuming.

Some other cool things:

• This tool works for a bunch of different languages.
• The definitions can be adjusted for different levels.
• The user can keep track of all previous definitions to later review them.

This has many benefits that help the user stay engaged with the content and remember new words. What are these benefits?

Why does inContext help you remember?

When I moved to Japan as a 24 year old, I thought I had a pretty good idea of how language acquisition worked. I had spoken two languages since childhood, and I thought I could approach learning another scientifically and systematically. It took just a few short days in rural Japan to realize that my ideas on language acquisition were wrong. I spent much of the next three years as a language teacher, and later language student trying new things and seeing what stuck. I was also fortunate to be able to observe my students, co-workers, and friends doing the same. As I researched and grew proficient in conversational Japanese, I defined a few mental models and distilled language acquisition into 3 key points:

1. Try to do things you are already interested in but in your target language. In other words, have a clear motivation for doing something beyond just learning the language. The most important thing in languages is to try to understand the message, and the stronger the reason for understanding that message, the faster you will learn. My roommate is a big fan of Spanish soccer, so he’s very motivated to read the news and listen to press conferences in Spanish. His strong interest in the content makes the language practice much easier, and he ends up consuming a lot more than he would otherwise
2. Learning a language is never easy, but it should be uncomfortable in the right way. I never found the act of shoveling as many words as possible into my brain using flashcards particularly effective; it was tedious and frustrating. But if I kept running into a particular word in a situation and was forced to slow down and puzzle out its meaning using context, then I often remembered it a lot better. This is where reading definitions in your target language comes in: it not only gives you more practice with the language, but forces you to engage with the meaning of the word in the context in which it belongs. There is significant research in psychology showing that processing a word at a deeper level will help it stick around much longer. If you do the work up front to internalize the contextual meaning of a word or phrase, then this struggle will be repaid by improved recall of this word or phrase.
3. Spend as much time as possible in a state of i + 1 comprehensible input. The idea of comprehensible input comes from linguist Stephen Krashen’s input hypothesis, which states that language is acquired when the learner understands messages slightly above their current level. As Krashen frames it, if i is your current level then the level slightly above yours would be i + 1. Any input that is too easy (i or i - 1) will not help you advance, and hard input (i + 10) will be impossible to comprehend and you’ll be likely to disengage. Historically, this has been quite difficult; you either need a tutor, a parent willing to speak simply, or you need things like graded readers.

None of these points are novel, they are just the results of my lived experience combined with the research and techniques that I read about. I designed inContext to actively address all 3 of the points described above:

1. inContext works on whatever you are reading on the internet. Presumably stuff you are already interested in!
2. By creating a definition in the target language, inContext provides the reader with the information necessary to puzzle out the meaning of the word. It’s up to the user to actively work to understand this meaning. This has two benefits: the user is more likely to remember the word since they are engaging deeply with its meaning, and the user gets a much better sense of the use of the word in context without reverting back into their more comfortable language.
3. The tool gives you definitions targeted to your language level, at roughly the i + 1 level. If you are at an intermediate Japanese level then reading a dictionary designed for native speakers of Japanese (i + 10 level) is not as helpful as reading a definition meant for someone at an upper-intermediate level. With AI we can now create definitions and tailor them to the learner’s i + 1 level, something that wasn’t possible before.

inContext isn’t designed to make language learning automatic or effortless. Rather, it provides you with the information you need to learn your target language the way humans actually learn languages.

Further Reading

In closing, I’ll also list a few of the valuable sources of information that served as the inspiration for this tool:

• Yomitan (formerly Yomichan) is an amazing (non-AI) dictionary tool, and an important inspiration for this app.
• An interview with linguist Stephen Krashen who defined the method of comprehensible input.
• A modern day retrospective on Krashen’s research.
• An interview with Steve Kaufman who has used this method to learn many languages.
• Why you should look up definitions in your target language, and other topics by the Japanese language learning community.

We are excited about how AI is improving language education, and we will keep looking for ways to make the language experience better and better.