Dictionary Information

Dictionary Information

The dictionary is the shared word list every game mode plays against. As of this writing it holds approximately 1.67 million entries, stored one word per line in the bot's main dictionary file. This page documents how words enter and leave it, how non-standard input gets converted into a valid entry, and the standards curators apply when reviewing a suggestion.

There is also an ideological vision of dictionary criteria known as Super-Shiritoriism — see the full write-up on the external wiki if you'd like to read it, or click here for the page on this wiki. That framing is stylized and in-universe, not a technical reference; this page documents the same underlying mechanisms in plain, verified terms instead.

For the full set of themed word lists that supplement the main dictionary (colors, countries, creatures, mythology, and dozens more), see List of Dictionaries.

What Counts as a Word?

Word curators work from a deliberately broad working definition of "word":

"A single distinct meaningful element of speech or writing, used with others (or sometimes alone) to form a sentence and typically shown with a space on either side when written or printed."

In practice, this covers:

• Acronyms and initialisms (e.g. NASA, FBI).
• Proper nouns with cultural or historical significance (e.g. Napoleon, New-York).
• Taxonomy names and classification numbers, if commonly used.
• Internet slang, emoticons, and expressive strings (e.g. UwU, '-w-', awa).
• Written forms of numbers and numeral-based media terms (e.g. one-percent, 480p → four-eighty-p).
• Foreign loanwords that have entered English usage (e.g. sushi, déjà-vu, siesta).

Special case: a song title is only accepted if the song itself can be found on YouTube.

Suggesting Words

Players propose new words with /suggest. Submissions are checked against the live dictionary before anything is posted: words already present are reported back separately and excluded, and a submission must contain between 1 and 1,001 new, valid words after that filtering (a request that resolves to zero new words is rejected outright). Word removal requests go through the same command and the same limits, routed to a separate review queue.

Every batch of new words is posted to a curator review channel (rejections/removals to a second, dedicated channel) with four options available to word curators:

• Accept — writes the words directly into a staging area, from which they merge into the live dictionary automatically about every 2 minutes.
• Review Manually — sends the curator a file of the submitted words by DM for closer inspection; the curator corrects the list by hand and applies it themselves rather than accepting it as-is.
• Reject — declines the submission outright.
• Cancel — available only to the original submitter, to withdraw their own pending request (for example, to fix a mistake) before a curator acts on it.

A user can be barred from submitting suggestions entirely; attempting /suggest while banned returns a rejection message instead of posting anything for review.

Normalization (Dictionary Content)

The dictionary only stores words made up of the Latin letters A–Z, the apostrophe, and the hyphen. Loanwords and other non-standard input are accommodated into that form following ordinary standard-English romanization: a borrowed word keeps its accepted English spelling, with diacritics simply dropped rather than substituted for anything else. For example: café → cafe, résumé → resume, touché → touche, façade → facade, jalapeño → jalapeno, über → uber.

Numerals are not permitted in written form at all; a number must be spelled out in words, joined by hyphens rather than spaces (see Hyper-Hyphenation below) to form a single valid entry — for example, 1345 is only valid as one-thousand-three-hundred-forty-five. See Grammar and Form below for how a number is chosen to be spelled once it's accepted.

This is the editorial convention curators apply to decide what a word is actually stored as. It's a separate matter from how the bot's live gameplay engine converts a player's typed input into a matching key while a game is in progress — that mechanism follows different, code-defined rules of its own (including some that don't simply drop diacritics the way the convention above does) and is documented on its own page: Word Normalization.

Curator Guidelines

The rules below are the standards word curators apply when reviewing a suggestion. None of them are enforced mechanically by the bot at submission time — accepting a word as a legitimate entry is a human judgment call curators make against these standards.

Formatting

• Diacritics are accommodated by dropping them, per standard English loanword convention (see Normalization above) — e.g. jalapeño is stored as jalapeno.
• Acronyms and abbreviations may be added alongside their long form, joined into a single hyphenated entry (e.g. png alongside papua-new-guinea).
• Phrasal words allow each individual word of the phrase to be suggested separately as its own entry (e.g. molotov cocktail allows both molotov and cocktail individually), except where the phrase is a non-English proper noun, or a proper noun sourced only from user-generated content.
• Nouns, verbs, and (where attested) adjectives may be suggested together with their inflected forms: most nouns take -s (stove → stoves); verbs take -ed, -s, and -ing (judge → judged, judges, judging); adjectives take -er/-est where attested (silly → sillier, silliest). A proper noun may fit any of these categories, or function as its own part of speech.
• A leading or trailing hyphen on a submitted word is ignored rather than treated as part of the word.
• An entry may not exceed 450 characters.

Grammar and Form

• Possessives follow standard English rules: a word already ending in s takes only a trailing apostrophe (e.g. boss → boss'); any other word takes 's (e.g. teacher → teacher's). In practice, this rule is not applied to possessive forms of song titles, even though nothing in the rule itself excludes them — a song title is a proper noun like any other and would qualify under the same standard. This gap traces back to a specific misreading of what counts as a "proper noun phrase" by the Discord Bot's owner, Ghost; it has not been formally corrected, so treat possessive song titles as unsupported in current practice regardless of what the rule as written would suggest.
• Plurality must match how the word is actually used, not be added reflexively.
• Spelling should follow standard or otherwise widely accepted forms.
• Numbers are spelled the way they're actually pronounced, not by full place-value expansion — e.g. 480p is four-eighty-p, not four-hundred-eighty-p.

Sourcing

• Discord messages, YouTube or streaming-platform comments, and individual forum posts (single comments, not the established resources described below) are not accepted as sources, since they can be edited or deleted without notice.
• Song lyrics are not considered a valid source on their own.
• AI chatbots are not a valid source under any circumstance — see AI Isn't a Valid Source below.

User-generated content

As of a policy update dated November 13, 2025, words sourced from user-generated content platforms — Urban Dictionary, Fandom/community wikis, dedicated technical wikis, and established forum FAQs — are accepted, but only under three mandatory conditions, all of which must be met:

1. Age: the word must have been verifiably public on its source platform for at least 180 days (6 months) before it's submitted. This is meant to filter out content invented for the submission itself rather than something that already existed and saw use.
2. Creator distance: the word's original creator, or anyone connected to that creator, cannot be the one who submits it — including asking someone else to submit it on their behalf. This is an intentional, acknowledged bias: it exists specifically to require that a word has been adopted by people outside its creator's own circle before it's considered, rather than accepting a creator's (or their friends') own word on their say-so.
3. Source type: the word has to come from a recognized group community resource, not an unreliable low-credibility source like a single blog comment, an exaggerated one-off Reddit post, or an unmaintained personal site.

Any submission curators consider "sus" under these rules gets a request for the source link and an explanation of the word; if a curator can't quickly verify the word's age or that the submitter isn't its creator, it's rejected outright rather than given the benefit of the doubt.

Content

• Complete gibberish is not accepted.
• Words no longer used in modern English are not accepted, except for archaisms and obsolete words that are still recognized by standard dictionaries.
• Words containing an em dash are not addable.
• Unintentional typos are not accepted, unless the misspelling is itself part of an established proper noun phrase.
• Numbers are restricted: any number-containing word without an alternate name, whose value is not between −1 and 101 (i.e. 0 through 100), and which is not a standardized value under some recognized instrument or counting system, is invalid. Separately, if the specific number in a word could realistically be swapped for any other numeric value without the word losing its meaning, it is not accepted — this excludes essentially arbitrary numeric strings dressed up as words.

Curator Conduct

• A curator cannot approve their own submission — if a curator uses /suggest themselves, a different curator has to review it. In practice, this rule has fallen out of active use: curators now mainly add their own words directly with !a (see Curator Commands below) rather than routing them through /suggest, which sidesteps the approval step this rule was written for. Treat it as outdated courtesy rather than enforced practice.
• When a suggested word is obscure or its meaning isn't self-evident, suggesters are asked to attach a short clarifier so curators don't have to research it blind, for example: ^ musician, ^ book title, ^ slang, ^ city, or ^ (link).

Curator Commands

Curators can bypass the /suggest review queue entirely and edit a dictionary file directly:

• !a word dict — add a word.
• !r word dict — remove a word.

If no dictionary is named, the command targets the main dictionary. Neither command accepts multiple words in one message — a batch add or remove has to be submitted as a file instead.

Hyper-Hyphenation

Typing !hyphens, !hyper-hyphenation, or !wysiwyg in a game channel brings up the bot's own explanation of this rule:

Hyper-Hyphenation (noun)

1. (linguistics, normalization) Joining two or more separately-attested words into a single hyphenated expression, treating them as one unified lexical item.

2. Super-Shiritori's normalization rule: when an abbreviation, acronym, initialism, or multi-word number is expanded into its full multi-word form, those component words get joined with hyphens into ONE valid dictionary entry; e.g. CAPTCHA becomes completely-automated-public-turing-tests-to-tell-computers-and-humans-apart.

3. (by extension) Hyphenation is the mechanism, not the justification. An expansion still has to be independently real/attested; Hyper-Hyphenation only governs how its already-real words get stitched together, never whether they belong.

In practice: an acronym's expanded form is valid alongside the acronym itself, provided that expansion is genuinely attested (a real, documented long form) rather than invented to make an otherwise-invalid string of letters pass as a word.

To see the definition of Hyper-Hyphenation, go to the Word Compendium page. That framing is stylized and in-universe, not a technical reference; this page documents the same underlying mechanism in plain, verified terms instead.

Neosesquipedalianism

Typing !neoses or !neosesquip brings up the bot's explanation of the opposite, invalid, pattern:

Neosesquipedalianism is the idea that you can make a "new" word by stripping the spaces, hyphens, and apostrophes out of an already-existing multi-word term and mashing what's left into one unbroken string.

Example: the mineral Clino-ferro-ferri-holmquistite becomes Clinoferroferriholmquistite. Nothing about the word actually changed except the punctuation holding its real components apart; it's the same term with its formatting deleted, not a new one.

This is not valid here. It's the mirror image of Hyper-Hyphenation (!hyphens): Hyper-Hyphenation legitimately joins independently-attested words together with hyphens; Neosesquipedalianism just erases the punctuation of an existing compound and calls the result new. Removing separators isn't creating a word; it's deleting information.

AI Isn't a Valid Source

Typing !ai, !noai, or !aisource brings up the bot's standing position on AI-sourced suggestions:

Asking an AI chatbot whether something is a word, or having it "confirm" a definition, isn't an acceptable source for adding a word to the dictionary. This includes ChatGPT, Claude, Google Gemini, Grok, DeepSeek, Microsoft Copilot, Meta AI, Perplexity, and any other popular AI model or chatbot.

AI models hallucinate; they'll state a word, spelling, or definition with total confidence even when it doesn't actually exist anywhere. That confidence doesn't make it real, and there's nothing to independently check it back against.

Words need an attestable, citable source: a real dictionary, an established compound, a documented coinage; something someone else can actually go look up and verify. "I asked an AI and it said yes" isn't that.

See also

• List of Dictionaries
• Word Normalization