Data Sources

Search dictionaries

The full search list is generated from the word-list npm package by Sindre Sorhus. The installed package is MIT-licensed and points to the English word list maintained in the atebits Words project. WordTorch cleans and normalizes that source, limits entries to supported lengths, and removes locally blocked terms before writing the static dictionary files used by the tools.

The current generated full dictionary contains 269,865 entries from 2 to 15 letters. A separate five-letter set contains 12,578 entries for the five-letter tools.

Common Words

The Common Words subset starts with frequency data from the MIT-licensed @derock.ir/words-frequency npm package, which describes its source as common words ranked from Project Gutenberg text. WordTorch keeps cleaned entries that also occur in the full dictionary and adds a small reviewed override list for obvious everyday forms missed by the cutoff.

The current Common Words set contains 12,080 entries, including 80 reviewed overrides. Frequency is a practical sorting signal, not a definition of correctness. A familiar regional word can be absent from Common Words while still appearing in All Words.

Definitions and related words

Some definitions, synonyms, antonyms, and example sentences are generated from Princeton WordNet 3.1 data distributed through the wordnet-db npm package. WordNet is a lexical database: it groups distinct senses of a word and records relationships between those senses.

WordTorch generates static detail chunks, filters unsuitable display material, and uses a small set of reviewed overrides where an everyday word would otherwise receive an unhelpful first sense. WordNet does not cover every spelling in the larger game dictionary, so some valid search results have no detail panel.

WordNet is copyright Princeton University and is used under the Princeton WordNet license. Read the complete Princeton WordNet license notice.

How the data reaches the tools

The source packages are development inputs. Build scripts generate local text files and small letter-based JSON chunks, which the deployed static site loads only when a search needs them. WordTorch does not call a remote dictionary service for each query.

The same core lists power letter counting, exact anagrams, prefixes, suffixes, contained groups, crossword positions, and color-clue filtering. The score calculator uses a separate fixed table of standard English-language tile values; its dictionary message is only a local presence check.

Limits and game dictionaries

No general word list matches every crossword publisher, classroom list, board-game edition, or online word game. Some games permit inflections or regional spellings that others reject. Names, abbreviations, multi-word phrases, and themed entries may also fall outside the local dictionaries.

WordTorch results are therefore candidates for solving and learning. When acceptance affects a score or competitive play, the official word list for that game or puzzle is the final authority.