Parser V2 Research Notes
Parser V2 is an earlier private grammar-layer research direction for ViaSign.
It is designed to help prepare English input for sign language accessibility work, beginning with Singapore Sign Language exploration.
The purpose of Parser V2 is not to produce final signing by itself.
The purpose is to understand sentence structure, identify important grammar signals, and prepare information that can later support motion, review, and signing output.
Current Public Boundary
The published viasign-parser does not currently implement the SgSL grammar signals or motion-ready information described below. Its public contract is limited to closed natural-English surface categories, visible uncertainty, review_required=true, and motion_ready=false.
The items in this document are research notes, not shipped capability, benchmark evidence, or SgSL authority. They must be revalidated against the current implementation and reviewed with relevant community knowledge before any future public claim.
Why a Parser Layer Exists
Sign language AI should not begin only with an avatar.
Before movement, there needs to be meaning.
Before animation, there needs to be structure.
Before output, there needs to be review.
A parser layer helps separate these concerns.
Instead of asking the system to immediately generate a signing output, Parser V2 asks questions such as:
- What kind of sentence is this?
- Is this a question?
- Is there negation?
- Is there a time marker?
- Is there a location reference?
- Is there a topic-comment structure?
- Is there a reference to a person, object, or place?
- Does this sentence need signing space, movement, or human review?
This makes the system easier to inspect, test, and improve.
What Parser V2 Looks For
Parser V2 focuses on grammar and structure signals that may matter for signed language processing.
Examples include:
- question patterns
- negation patterns
- time markers
- location questions
- topic-comment structure
- completion markers
- rhetorical question patterns
- references to people, objects, and places
- source-gloss indexing
- motion-readiness flags
- review-required flags
These are early engineering signals, not final language authority.
The parser is meant to support structured exploration and review, not replace Deaf fluency or community knowledge.
Grammar Before Motion
Parser V2 follows one important principle:
Grammar before motion.
This means the system should first understand the sentence before attempting to generate motion.
For sign language accessibility, this matters because signed languages are visual, spatial, expressive, and grammatical.
A sentence is not simply a chain of signs.
Some meaning may need facial expression.
Some meaning may need signing space.
Some meaning may need directional movement.
Some meaning may need topic setup before comment.
Some meaning may need review before it can be trusted.
If the grammar layer is weak, the motion layer will also be weak.
Review Flags
Parser V2 is designed to keep review visible.
For early sign language AI work, it is important that the system does not pretend to be certain when it is not.
Review flags help mark areas that may need human checking.
This is especially important for signed languages because a small change in movement, space, timing, or expression can affect meaning.
A review-first system should make uncertainty visible instead of hiding it behind polished output.
Motion-Ready Information
Parser V2 can prepare information that may later support a motion layer.
This does not mean the parser generates final motion.
It means the parser can identify parts of a sentence that may be relevant to motion planning, such as:
- whether a sentence is a question
- whether negation is present
- whether a time marker should be established
- whether references may need signing space
- whether motion or direction may be involved
- whether the sentence should be reviewed before output
This creates a bridge between grammar understanding and future motion generation.
What Parser V2 Is Not
Parser V2 is not a complete Singapore Sign Language translator.
It is not a replacement for Deaf teachers, interpreters, signers, or community reviewers.
It is not a claim that AI output is automatically correct.
It is not the final motion engine.
It is not the final avatar layer.
Parser V2 is an early grammar and structure layer that helps ViaSign move carefully toward more auditable sign language accessibility tools.
Engineering Direction
The long-term direction is to keep the system layered and auditable.
A possible flow is:
- Input sentence
- Parser analysis
- Grammar and structure signals
- Review flags
- Motion-ready information
- Motion planning
- Human or community review
- Future signing output
This structure helps keep the work understandable.
It also makes it easier to test each layer before moving to the next.
Guiding Principle
Parser V2 exists because ViaSign should not rush from text directly into animation.
The system should first ask what the sentence means, what structure it has, and what parts require care.
The goal is not only to generate output.
The goal is to build a system that can be checked, reviewed, corrected, and improved over time.
That is why Parser V2 is part of ViaSign’s foundation.