Next Biz Thing #421 kotomora.com
Kotomora https://kotomora.com
Markus J. Diplama looks at Kotomora, a Japanese dictionary and grammar reference built in Tokyo by a team of language learners working alongside native speakers. The episode explores what sits behind its dictionary of more than 216,000 words: a sentence analyzer, 862 JLPT graded grammar patterns, kanji stroke order, pitch accent and translation guides that compare similar words side by side. It makes the case for treating a language as a connected map rather than a flat list of definitions, and for teaching nuance and register instead of handing over a verdict. Worth a listen for anyone learning Japanese, or building learning tools.
Have you ever looked up a word in another language, found what looked like the perfect translation, used it in a real conversation, and then watched the other person's face do something you did not quite expect? Not confusion exactly. Something gentler than that. A small flicker that says, I understand you, and also, that is not quite how we would have said it. If you have ever had that moment, you already understand the problem that today's business is built around, and you already know that it is a much bigger problem than most dictionaries are willing to admit.
Hello and welcome back to The Next Biz Thing. I am Markus J. Diplama, and every episode I go looking for the companies that are quietly fixing something the rest of us have simply learned to live with. Not the loud launches, not the companies that exist mostly to be talked about, but the ones building something careful and useful and then getting on with it. Today I want to talk about a company working in a field that is extraordinarily crowded and yet, strangely, still full of gaps. Language learning. And specifically, a Japanese dictionary and grammar reference called Kotomora.
Kotomora was built in Tokyo by a team of language learners working alongside native Japanese speakers, and I want to sit on that detail for a second because I think it explains almost everything that follows. This is not a product designed by people who finished learning Japanese and then forgot what the hard parts felt like. It was made by people who were in it, who kept running into the same wall, and who had the good sense to build the thing they wished existed rather than just complaining about the thing that did.
The wall, as they describe it, is the gap between a dictionary definition and real world usage. Every learner meets this wall eventually. You look up an English word and the dictionary hands you back four Japanese options with no guidance about which one a real person would actually reach for. They all mean the thing. They do not all mean the thing in the same room, to the same person, at the same moment. One is warm, one is formal, one is what you would write but never say, and one would make a colleague think you had learned the language entirely from television. The dictionary is not wrong. It is just not finished.
So Kotomora set out to finish it. Their whole approach is built around explaining how words and patterns are actually used, which means nuance, register, and context, rather than simply handing over a gloss and wishing you luck. That sounds like a small shift in emphasis. In practice it changes what the product has to be.
Let me walk you through what they have built, because the scope of it genuinely surprised me. At the centre there is a dictionary with more than two hundred and sixteen thousand searchable words. That alone would be a serious reference. But around it sits a set of tools that do the work a dictionary normally leaves to you. There is a sentence analyzer that takes a complete Japanese sentence and breaks it apart, giving you the readings and showing you the grammar holding it together. If you have ever stared at a wall of Japanese text knowing every individual word and still having no idea what the sentence was doing, you will understand why that tool matters.
There are eight hundred and sixty two grammar patterns, organised by JLPT level from N5 all the way up to N1. That organisation is not cosmetic. It means a beginner is not drowning in structures they will not need for two years, and an advanced learner can find the specific pattern they are circling without wading back through the basics. There is a full kanji reference, including stroke order diagrams, which are drawn from the openly licensed KanjiVG project. There are translation guides that take similar English words and lay out their Japanese equivalents side by side, showing you where each one actually belongs. And there are word lists organised by topic, by level, and by register, which is that last word again, register, doing a lot of quiet work.
Then there is the detail on an individual entry, and this is where you can see the philosophy showing through. A word does not just come with a meaning. It comes with its reading, with furigana, with pitch accent, with conjugation, with its JLPT level, with how the grammar around it forms, with real example sentences, with audio, and with the related patterns you are likely to meet next. Pitch accent in particular is one of those things that separates a learner who is understood from a learner who is understood immediately, and it is routinely left out of resources aimed at beginners because it is fiddly. Including it is a choice about respecting the learner.
But the thing I find most interesting about Kotomora is structural rather than featural. Everything is clickable, and everything connects. Words, kanji, grammar points, example sentences, all of it links through to everything else. The team describes the result as a connected map rather than a set of isolated lookups, and I think that framing is exactly right. Because the truth about language is that it is not a list. It is a web. Words lean on each other. A grammar point only makes sense once you have seen it carrying real sentences. A kanji stops being a shape and starts being a meaning once you have met it in six different words. A reference tool that lets you follow those threads is doing something a flat dictionary structurally cannot do, no matter how many entries it holds.
Now, the other thing worth talking about here is where the underlying data comes from, because this is a part of the business that a lot of companies would keep quiet and Kotomora is refreshingly open about. They build on established, openly licensed resources: the JMdict and EDICT dictionary projects and the KANJIDIC kanji database, both from the Electronic Dictionary Research and Development Group, along with the Creative Commons licensed KanjiVG stroke order diagrams. And then they review the content with native speakers on top of that foundation.
I want to make the case for why that is a strength rather than a shortcut, because I think some listeners will hear open data and think of it as borrowed. It is the opposite. Those projects represent decades of accumulated, carefully maintained lexical scholarship. Trying to rebuild that from scratch as a startup would not produce something better, it would produce something thinner and more error prone, and it would take years you could have spent on the part that is actually missing. Kotomora's judgement call was to stand on the authoritative data where authoritative data exists, and to spend their own effort on the layer nobody had built: the usage, the comparisons, the context, the connections. That is a mature decision. It is the decision of a team that is interested in the learner's outcome rather than in the size of their own moat.
And they have carried the whole thing onto mobile, with apps for iOS and Android, which matters more than it sounds. Because the moments when you most need this tool are not the moments when you are sitting at a desk. They are the moments in a shop, on a train, mid conversation, reading a sign, when a word arrives and you have about fifteen seconds of curiosity before it evaporates. A reference that lives in your pocket catches those moments. A reference that lives on a laptop mostly does not.
So why does any of this matter beyond people studying Japanese? I think it matters because it is a small, clear example of a shift happening right across education technology. For a long time the assumption was that learning was fundamentally a memory problem, and so the tools were memory tools: flashcards, drills, lists, streaks. Those things work, up to a point. But the point where they stop working is exactly the point learners get stuck, because the next thing you need is not more items in your head, it is a sense of how those items behave. Judgement. Feel. The ability to choose between four correct answers and pick the one that fits the room.
That is a much harder thing to teach, and it is why most tools do not try. Kotomora is trying, and it is trying by giving learners the context rather than the verdict. Not, here is the right word. Instead, here is what each of these words is doing, here is who says them and when, now you decide. That is a genuinely respectful model of teaching, and it produces a different kind of learner at the end of it: one who can handle a sentence they have never seen before, rather than one who can only handle the sentences they have already memorised.
There is something else I appreciate about this company, and it is the modesty of it. Kotomora is not promising fluency in thirty days. It is not gamifying your self worth or sending you guilt trips about your streak. It is a reference tool, built carefully, by people who use it themselves, aimed at the long unglamorous middle of learning a language where most of the actual progress happens. In a market full of noise, building something quiet and genuinely useful is an underrated act.
So if you are learning Japanese, or you are thinking about starting, or you have one of those half finished attempts sitting in a drawer somewhere quietly judging you, this is worth a look. Go and search a word you think you already know, and then follow the links outward and see where the map takes you. And if you are not learning Japanese at all, I would still encourage you to look at how this thing is put together, because the design thinking in it, the idea of building a connected map instead of a list, applies to a lot more than language.
That is Kotomora, out of Tokyo, doing the patient work of explaining not just what words mean but how they live. Thank you for spending this time with me. If this episode gave you something useful, share it with someone who is midway through learning something hard. I am Markus J. Diplama, this has been The Next Biz Thing, and I will see you next time.
