Skip to content

Method: word lists, licences and how the search works

Which open-licence word lists Lettercraft uses, what was filtered out, the licence texts, and exactly how anagram, wildcard and clue matching work.

Word lists and licences

English: SCOWL, the English Speller Database (ESDB) by Kevin Atkinson, https://wordlist.aspell.net/. Copyright 2000-2026 Kevin Atkinson. Licence: permission to use, copy, modify, distribute and sell word lists created from the database is granted without fee, provided the copyright notice and this permission notice appear in all copies and in supporting documentation. The full notice is in /words/en/LICENSE.txt. Most of the underlying data is in the public domain, mainly 12dicts and ENABLE2K.

Swedish: SALDO's morphology (saldom.xml, updated 2017-09-19) from Språkbanken Text, University of Gothenburg. Licence: Creative Commons Attribution 4.0 International (CC BY 4.0), https://creativecommons.org/licenses/by/4.0/. DOI: https://doi.org/10.23695/agcm-ny22.

Attribution: Språkbanken (2017). SALDO's morphology. Enriched and distributed by Språkbanken. We filtered the lexicon's written word forms, sorted them and split them by word length. The derived list is shared under the same licence, CC BY 4.0. The text is in /words/sv/LICENSE.txt.

The name and licence of the list in use are shown under every result and in the footer.

What was kept

The English list has 81,072 words, 5,185 of them with five letters. The Swedish list has 398,800 words, 9,543 with five letters. Both are split into one file per word length, 2 to 15 letters.

  • English: lowercase a–z only, so proper nouns, acronyms and possessives are dropped. SCOWL size 60, US and GB(-ise) spelling, default variant level, no special lists, diacritics stripped.
  • Swedish: base forms and inflected forms of nouns, verbs, adjectives, adverbs, prepositions, pronouns, conjunctions, subjunctions, interjections and the article. Proper nouns, abbreviations, multi-word entries, compound-prefix forms and genitive forms are dropped.
  • Swedish words may only contain a–z, å, ä and ö. Words with other letters, such as é, are left out.
  • Both languages: words without a vowel are removed, except a short list of real words and interjections such as hmm and brr. That clears unit abbreviations such as kg and km.

Known limits of the lists

These are spell-checker lists, not tournament dictionaries. A few abbreviations that contain a vowel remain in the English list, for example etc, esp and dept. The Swedish list comes from a lexicon that also records rare, old and dialect forms, so it is broader than the Swedish Academy word list.

A word can therefore be here and missing from your game's dictionary, or the other way round. Check a disputed word in the dictionary your game uses.

Anagrams

Two words are anagrams exactly when their letters match once sorted. The tool counts how often each letter occurs in the input and in each word of the same length. A ? fills one position: the other letters of the word must be exactly the letters you typed. That is why 15 letters cost almost as little as 5.

Word finder with tiles and blanks

A word fits when it can be spelled from your tiles with each tile used at most once. Where letters are missing, the number of missing letters must not exceed the number of blanks (?). Letters typed in Starts with, Ends with and Contains count as tiles already on the board and are taken off the need first.

Wildcard patterns

The pattern is checked character by character and only letters, ?, _, ., *, and bracketed sets with an optional ^ are let through. It is then translated to an anchored regular expression: ? becomes one letter, * becomes any run of letters and [ ] becomes a character class. Nothing you type is ever read as expression syntax, and a pattern can fix at most 15 letters.

Clues for five-letter games

Greens are settled first. Each remaining letter of the answer can then turn at most one tile yellow, left to right. The helper keeps the words that would have produced exactly your colours.

Suggestions are ranked by the expected number of words left when 400 candidates or fewer remain, and by letter frequency among the candidates above that.

Privacy and speed

Everything runs in your browser. What you type is not sent to a server. Word files are fetched per word length when needed and kept for the visit. Shareable links store your input in the address, not on a server.

Tests

Anagrams, the word finder, wildcard patterns and clue matching are tested against the shipped lists. Each test file names the source and licence of the words it uses, and a separate test checks that the licence files and attributions are present.

Sources