Two different jobs
A translation app and a speaking practice tool are not competitors, because they are aimed at different outcomes. One completes a transaction now; the other changes what you can do next month.
Confusion arises because both appear to be about communicating in Japanese. In practice, using an app is closer to using a lift than to climbing stairs, and which one you want depends entirely on whether you are trying to get somewhere or get stronger.
| Need | Translation app | Practice |
|---|---|---|
| Understand this sign now | Ideal | Irrelevant |
| Complete a transaction today | Ideal | Slower |
| Handle it yourself next time | No effect | The point |
| Read a contract | Useful, with care | Not sufficient |
| Talk to a colleague | Awkward | The point |
| Emergency | Ideal | Do not experiment |
The last row deserves stating plainly. When something matters urgently, use whatever works and learn from it afterwards.
How apps quietly stall progress
The mechanism is specific. Memory for words is built largely by retrieval effort, and reaching for a phone removes the effort before it happens. The word is supplied, the sentence completes, and nothing is retained.
Repeated over months, this produces a learner who has looked up the same twenty words hundreds of times and knows none of them, which is a common and demoralising position.
The second effect is subtler: an app lets you avoid ever discovering your actual gaps, because every gap is filled instantly. Your sense of what you can say stops matching reality.
- Looking up before attempting prevents retrieval.
- The same words get looked up indefinitely.
- Gaps stay invisible because they are always filled.
- Speaking confidence does not develop.
- The app becomes the thing you are fluent in.
Using one so it teaches
The fix is a sequence rather than abstinence. Attempt first, however badly, then check, then say the corrected version aloud, then note it.
That order preserves the retrieval effort and adds the correction on top of it, which is roughly the ideal condition for learning a word.
It takes about fifteen seconds longer than looking straight up, and it is the difference between a lookup and a lesson.
| Japanese | Romaji | Meaning |
|---|---|---|
| 何と言えばいいですか。 | Nan to ieba ii desu ka. | What should I say? |
| これは日本語で何ですか。 | Kore wa Nihongo de nan desu ka. | What is this in Japanese? |
| 合っていますか。 | Atte imasu ka. | Is it correct? |
| もっと自然な言い方は？ | Motto shizen na iikata wa? | A more natural way to say it? |
| 書いてもらえますか。 | Kaite moraemasu ka. | Could you write it down? |
| 自分で言ってみます。 | Jibun de itte mimasu. | Let me try saying it myself. |
The last phrase is the habit worth building with real people too. Most will happily wait a few seconds while you try.
Where machine translation goes wrong in Japanese
Japanese is unusually hostile to translation without context. Subjects are omitted, politeness level shifts with the relationship, and the same sentence can be appropriate or rude depending on who is speaking to whom.
A translator sees a sentence with none of that, so it guesses, and the guesses tend toward a neutral middle register that is wrong in both directions: too stiff among friends and too casual with a client.
This is why machine-translated Japanese often reads as correct but odd, and why using it for anything relational rather than transactional is risky.
| Feature | Why it breaks translation |
|---|---|
| Omitted subjects | The translator guesses who acted |
| Politeness levels | No relationship information available |
| Inside and outside groups | Direction of honorifics is invisible |
| Context across sentences | Each sentence handled alone |
| Implication over statement | Indirect refusals translate as agreement |
What practice does instead
The alternative is not memorising more; it is building retrieval speed for the things you actually say, which is what conversation practice does.
In Unihongo's 3D classroom the mechanism is deliberate: you attempt the sentence, fail, get the natural version, say it back, and the word you failed on goes into a deck built from your own gaps. That is the attempt-then-correct sequence, automated, which is exactly what a translation app skips.
Over a few weeks the effect is visible in a specific way: the situations where you reach for your phone get narrower.
- Attempt before receiving the answer.
- Say the corrected version aloud immediately.
- Review the words you failed on, not a generic list.
- Re-enter the same situation until the phone is unnecessary.
When to just use the app
Purism here is counterproductive. There are situations where using a translation app is simply correct, and feeling guilty about them helps nobody.
Official documents, medical situations, anything with legal or financial consequence, emergencies, and long written text all fall into this category. So does any moment where you are tired and the goal is to get home.
The distinction worth holding is between situations where you are trying to accomplish something and situations where you are trying to improve. Both are legitimate; they just call for different tools.
A reasonable rule: attempt it yourself when the cost of failing is mild, and use the app when it is not.