MemeLingua Say more with less.
A tiny language you can write with emoji and say out loud. Learn a small set of basic ideas, then combine them to make useful messages like I want food, help me, go there, or if you see fire, leave.
I want food.
tu aid mi
If you see fire, help me.
music, height, cover
Emoji, sound, meaning
Each basic idea has an emoji, a word you can say, and a meaning it keeps across sentences. The emoji helps you remember it; the spoken form lets the language work off-screen.
SELF
I or me. The speaker in a clause.
YOU
The listener or addressed person.
WANT
Desire. Kept separate from need and can.
NEED
Requirement. Stronger than want.
TO
Destination, recipient, or target.
FROM
Source, origin, or movement away.
ABOUT
Topic. Keeps speech and thought directed.
CHOOSE
Decision or selection. Not the same as yes.
Hear it spoken
Each clip says the MemeLingua words under the emoji, so you can hear how the short spoken roots connect.
Help me.
If you see fire, help me.
After done, I celebrate.
The person fears fire.
Speak about music.
How you make sentences
MemeLingua keeps the grammar visible. Put ideas in order, use relation words like to and from, and use a new line when one thought frames another.
Put ideas in order
A sentence starts with who or what, then says what happens, then adds the thing it happens to.
Point between things
Relation roots attach the thing before them to the thing after them.
Use a new line
A line break separates a condition, time frame, or short state from the clause it frames.
Make new ideas from small pieces
You do not need a separate word for everything. Put two familiar pieces together and the new idea is usually readable.
Music
Sound plus content. A thing you can hear as structured material.
Height
Measure plus on/up. Vertical measurement without a separate ladder root.
Cover
Container plus on/top. A lid as the top of a container.
Drink
Mouth plus liquid. Action and substance stay visible.
What makes it useful
The point is not to replace natural language. It is a small shared layer for quick messages, labels, instructions, games, and teaching.
Travel and shared spaces
Short signs, labels, warnings, and requests that need more structure than a single icon can carry.
Field instructions
Compact directions for help, movement, tools, source, destination, urgency, and completion.
Games and stories
A lightweight grammar for quests, puzzles, rules, dialogue, item behavior, and plot summaries.
Teaching grammar
Small examples make grammar visible: action, place, cause, yes/no, if, not, to, and from.
Say it out loud
Every emoji has a short spoken word, so a message can be typed, read, or spoken.
AI and tools
Short messages can be checked by software, searched, or turned back into normal language.
Tiny plot mode
One fun test is whether a tiny language can still carry the shape of a familiar story. These are deliberately TLDR, not full translations.
Back to the Future
A teen is thrown into the past, helps his parents fall for each other, then races back to his own time.
Harry Potter
A lonely kid discovers magic, finds friends, and faces the dark power that hurt his family.
Star Wars
A farm kid joins rebels, rescues a leader, and destroys the empire's giant weapon.
The Matrix
A hacker learns his world is false, chooses the truth, and fights the machines controlling people.
We test whether it still makes sense
A tiny language only works if the message survives. The repo includes practical tests and standard language-coverage checks.
Does it actually work?
The current draft uses 100 basic ideas. Numbers stay as ordinary numbers. The strongest results are in short messages, relation tracking, and very basic human meanings.
Example messages
Checks whether short MemeLingua messages decode back into the intended English.
Meaning maps
Current local score for source, destination, cause, topic, condition, and object cases.
Round trips
Starts in English, turns it into MemeLingua, then turns it back to see what survived.
Basic meaning coverage
Current strong coverage against a standard list of very basic human meanings.
Everyday vocabulary
Strong coverage on the Leipzig-Jakarta 100 list, a common basic-word benchmark.
Classic word list
Strong coverage on the Swadesh 100 list, using direct words, compounds, and numbers.
Start here
The source files are intentionally plain. The sheet is for scanning, the spec is for rules, and the eval docs show what the current design can and cannot do.
Vocabulary Sheet
Printable dark-mode table with number, emoji, meaning, spoken root, and origin.
docs/vocabulary-sheet.htmlCanonical Roots
The 100 single-emoji roots plus supplemental composition examples.
docs/vocabulary.mdSpoken Roots
Pronunciation layer with short roots and source rationales.
docs/spoken-roots.mdSpec
Surface form, clause order, relations, compounds, time, and doubling.
docs/spec.mdExamples
Working sentence examples in emoji form and source shortcode form.
docs/examples.mdEval Plan
Interpretability, API route, batch runner, scoring, and revision rule.
docs/eval-plan.mdLanguage Benchmarks
Leipzig-Jakarta, Swadesh, ASJP, Dolgopolsky, and NSM coverage.
docs/language-evals.mdMeaning-Map Eval
Checks whether who did what, to whom, from where, why, and when survives.
docs/amr-smatch-eval.mdMethodology
Root selection principles, benchmarks, roadmap, and next research steps.
docs/methodology.md