Katakana

Japanese is not written with a single character set, but with three side by side. One of these is Katakana. It consists of 46 basic characters with angular, straight strokes. Each character does not stand for a single sound like our “k” or “a”, but for an entire syllable: カ is pronounced “ka”, ト is pronounced “to”. Katakana is used above all when a word originates from a foreign language. “Computer” becomes コンピュータ in Japanese, pronounced roughly “konpyuuta”.

Why tech words in Japan look angular

Almost the entire vocabulary of the computer world is borrowed in Japanese. Software, byte, server, cloud, prompt: all these terms were reconstructed phonetically and then written in Katakana. Anyone who opens a Japanese news site about artificial intelligence will therefore see a striking number of these angular characters. They function like a visual signal: here comes a term from outside.

For computers, this is precisely a difficulty. A program that processes text must recognize that コンピュータ and the English “computer” mean the same thing. But the characters bear no resemblance whatsoever to Latin letters. Translation systems and search engines need their own rules and many training examples for this.

On top of that, the transcription is not unambiguous. Japanese has no “l” and no pure “v”. Both “light” and “right” become ライト. What is meant must be inferred from context. Humans manage this mostly effortlessly, machines find it harder.

46 characters, two strokes, and a small o

The system is manageably structured. There are five vowels plus rows made of consonant plus vowel: ka, ki, ku, ke, ko, and so on. This covers most Japanese syllables. Two small additional marks extend the set. Two strokes at the top right turn “ka” into “ga”, a small circle turns “ha” into “pa”.

For foreign sounds, combinations are used as a workaround. A normal-sized character followed by a smaller one produces a new syllable. This is how “fu” plus a small “a” creates the syllable “fa”, which does not actually exist in Japanese. Long vowels are written with a horizontal stroke, which is why so many loanwords end in ー.

It is important to distinguish Katakana from the other two scripts. Hiragana has the same syllables but looks round and flowing, and is used for Japanese words and grammar. Kanji are meaning-based characters adopted from Chinese, numbering in the thousands. A normal sentence mixes all three. A common misconception is that Katakana is a kind of alphabet for beginners. It is not an alphabet, because its unit is the syllable, not the individual sound.

From anime titles to character encoding in code

In everyday life, many people first encounter Katakana through pop culture. Anime titles, names of game series such as ポケモン (Pokémon), and product names of Japanese companies are often written in this script. Menus also use it for imported dishes. And advertising likes to use Katakana because its angular form comes across as modern and international.

In technology, the term comes up in character encoding. Unicode, the worldwide standard for numbering characters, has its own block for Katakana. There is even a narrow variant from early computing times, so-called half-width Katakana. To this day, it leads to display errors when systems confuse character sets.

For language models, Japanese is a good example of a fundamental problem. These systems break text down into small units called tokens. For languages without spaces between words, this breakdown is more difficult, and Japanese text often consumes more tokens than an English translation of the same content. Since costs are billed per token, using such models in Japanese tends to be more expensive.