Why Kinesema Exists

Update · 20 September 2026: The sentence-inspection preview described in this essay has been retired from the website. Parser development continues privately. See the current Kinesema page for what is available.

Kinesema exists because language access cannot be reduced to placing more text on systems designed around speech.

Captions, transcripts, and written information matter. I use and value them. But access through written English is not identical to access through a signed language. The difference is not only visual presentation. It reaches into how language carries structure, timing, space, expression, and meaning.

I began this work from my own experience as a Deaf and hard-of-hearing person in Singapore, and from one question that kept returning:

What would sign-language technology look like if language were part of the foundation instead of being added after everything else was built?

Kinesema is my attempt to keep following that question carefully.

Access is not one format

Accessibility is often described as though one alternative format can solve the whole problem. Add captions. Add a transcript. Add a sign-language avatar.

Each of those things can be useful. None of them becomes meaningful simply because it has been added to an interface.

Signed languages are not spoken languages acted out with the hands. They have their own grammatical structures and can use movement, location, timing, facial expression, gaze, and the space around the signer together. A system that replaces English words one by one may produce visible gestures while losing the relationships that carried the sentence’s meaning.

That is why Kinesema begins with Singapore Sign Language as a specific language context rather than treating “sign language” as one universal output mode. It does not claim to define SgSL or to turn general observations about signed languages into automatic SgSL rules. The current technical explanation keeps that distinction visible.

The goal is not to add motion to an English system. The goal is to understand what kinds of language, representation, evidence, and review would be needed before motion can responsibly express anything.

Structure before spectacle

An avatar is often the first thing people imagine when they hear “sign-language technology.” I understand why. Movement makes the idea visible immediately.

It can also make a weak system look finished.

Smooth motion cannot repair meaning that was lost before the hands began to move. If the language structure is wrong, a more polished avatar may only make the mistake harder to notice.

Before a system can move like language, it needs a reviewable account of the meaning and structure it is trying to express.

Kinesema therefore separates language analysis, proposed SgSL structure, human review, motion planning, and visual presentation into different stages. Each stage should expose what it received, what it changed, and what remains uncertain before the next stage is allowed to make the result look more complete.

This is the reason behind the principle language first, motion later. It is slower than beginning with an impressive avatar demo. It also gives the project a better chance of finding mistakes while they can still be understood and corrected.

The current system is deliberately small

The current protected Kinesema preview accepts an English sentence and exposes six bounded natural-English analysis fields beside one parser-owned candidate SgSL gloss. It keeps review_required: true and motion_ready: false visible.

That candidate is not accepted translation. It has not been approved as a natural SgSL expression, turned into a sign plan, or prepared for an avatar. The interface does not hide those missing stages behind a confident answer.

This makes the present system much smaller than the direction I am working toward. But the small boundary is useful. It tests whether an input can move through a private parser and return a public, inspectable result without shipping the private implementation into the browser. It gives people something concrete to question while keeping the unavailable layers visible.

Kinesema should grow only when the next layer has the evidence, representation, review, and public contract needed to support it. Capability should not be inferred from ambition, and a future roadmap should not be presented as though it already works.

Human communication is the point

My lived experience shapes why I notice this problem and why I continue working on it. It does not make me the authority for every Deaf person, signer, interpreter, teacher, learner, or SgSL user.

That is not a small disclaimer at the end of the project. It affects how the system must be built.

Technical tests can show that software followed a contract. They cannot decide whether a proposed expression is natural, culturally appropriate, or right for every context. AI can help organise research and implement systems, but it cannot manufacture community approval or replace people with relevant linguistic, cultural, accessibility, and lived knowledge.

Kinesema is intended to support people, not replace them. If it grows, I want it to become a bridge between language analysis, reviewable SgSL decisions, and careful motion—not a machine that asks people to accept whatever appears on the screen.

Success would not be an avatar that moves impressively. It would be a system whose important decisions can be understood, whose uncertainty remains visible, whose mistakes can be corrected, and whose direction is shaped by the people and language it is meant to serve.

Kinesema exists to make sign-language technology begin where language begins: with meaning, structure, and people.