How it works

What these dictionaries are made of, why they cover more than the alternatives, and how to get one onto your e-reader.

The synset pivot

A wordnet is a database that groups words by meaning rather than by spelling. Each group of synonyms — a synset — has an identifier. The important part is that every wordnet in the Open Multilingual Wordnet reuses the identifiers from Princeton WordNet, so the same concept has the same number in every language.

English   "wield"   ─┐
                       ├──▶  synset 01095899-v  ──┬──▶  "empunhar", "manejar"   Portuguese
Spanish   "empuñar"  ─┘                           └──▶  "impugnare"             Italian

That means the moment two languages are both mapped, a dictionary between them already exists. It has simply never been written down. Spanish to Portuguese needs no Spanish-Portuguese resource at all — only the two wordnets and a configuration file.

A generator that only reads Wiktionary cannot do this, because it needs a direct translation table to exist, and for most language pairs one does not.

Three layers

Why the inflection index matters

Books do not contain headwords. They contain wielded, geese, running. A dictionary that only indexes wield, goose and run will report “not found” for a word it actually knows. Every dictionary here ships a separate inflection index mapping hundreds of thousands of inflected forms back to their entry, and the coverage figures on each dictionary’s page count those forms.

Where these dictionaries fall short

Roughly four entries in ten have a definition but no translation: the wordnet knows the concept and your language has no single word for it. Those are shipped anyway, because a definition still helps, and they are counted separately in every metric rather than folded into the entry total. Only the pairs enriched with Wiktionary carry an inflection index — the form tables are the only bulk source of inflected forms, so a pair built from the wordnet pivot alone ships without one.

Installing a dictionary

KOReader reads StarDict dictionaries. Unzip the download and copy the folder — not its contents — into the dictionary directory for your device:

For the English-to-Portuguese dictionary on a Kindle, you should end up with a file at:

koreader/data/dict/en-pt-kdicts/en-pt-kdicts.ifo
koreader/data/dict/en-pt-kdicts/en-pt-kdicts.idx
koreader/data/dict/en-pt-kdicts/en-pt-kdicts.dict.dz
koreader/data/dict/en-pt-kdicts/en-pt-kdicts.syn

Restart KOReader. The dictionary appears under the dictionary list; long-press a dictionary’s name in a lookup popup to change the order in which several are consulted.

Since version 2020.04 you can put the dictionaries somewhere else by setting STARDICT_DATA_DIR = "/path/to/dict" in defaults.custom.lua.

If you have terminal access on the device, you can check the installation directly from the koreader/ folder:

sdcv -02 data/dict/ quaint

Getting KOReader onto a Kindle first

These dictionaries are for KOReader, not for the Kindle’s built-in reader. KOReader can be installed on any jailbroken Kindle: extract its release archive so that the koreader and extensions folders sit at the root of the Kindle’s USB drive, eject the device, and launch the new entry that appears in your library. The official guide covers each firmware version.

Jailbreaking is done at your own risk, and the official documentation is the authority on the procedure for your firmware — not this page.

Official documentation

• KOReader wiki — Dictionary support (installation paths, supported formats, testing)
• KOReader wiki — Installation on Kindle devices
• KOReader user guide
• Princeton WordNet
• Open Multilingual Wordnet
• Wiktionary
• kaikki.org — machine-readable Wiktionary extractions
• StarDict file format specification