Bio ORCID GitHub Zenodo Archive.org Part 0 (PDF) Part 1 (PDF) Press

Home Algebraic Foundation Visualizations Open Questions Sequence Visualizer

The Engineering of Alphabets (EOA) program investigates a class of deterministic symbolic representations derived from letter-name strings. The research spans symbolic dynamics, representation theory, deterministic systems, and the algebraic structure of phonetic encodings.

The recursive operator β. Schematic of the deterministic transformation that maps spoken letter names to numerical sequences.

What EOA Offers and Why It Is Unique

EOA is not a learned embedding. It is a structural primitive.

Training free and deterministic: No data, no loss function, no stochasticity, purely algebraic.
16 element sequence-group structure: The 26 English letters collapse to 16 unique sequences, an unexplained algebraic reduction.
Rank 1 dimensionality collapse: Word vectors built from these sequence groups collapse to a near one dimensional line.
Encoding dependent convergence: The Limiting Constant Ratio (LCR) shifts with phonetic encoding, from 2.710 (Hebrew) to 4.128 (Greek).
Algebraic foundation: 13 sequences form a true basis for a 13-dimensional solution space of a common linear recurrence.
Fractal geometry hints: Recursive, tree like branching underlies the sequence-group collapse, suggesting self similarity.
43 open questions: The program poses 43 open problems across 10 domains, requiring tools from mathematics, linguistics, and machine learning.

Core Observations

Convergence: Recursive operator β maps letter names to convergent numerical sequences.
Sequence-group collapse: 26 English letters collapse to 16 distinct sequence groups.
Rank-1 dimensionality: Word vectors built from these sequence groups collapse to near one-dimensional line.
Self-similar structure: Recursive, tree-like fractal geometry hinted at underlying the sequence-group collapse.

Limiting Constant Ratio (LCR): Encoding-dependent. See table below.

Encoding	LCR (λ)
English primary	3.306
English modified #1	3.037
English modified #2	3.173
English modified #3	2.689
English modified #4	3.209
English modified #5	3.141
Arabic	2.751
Hebrew	2.710
Greek	4.128

Limiting Constant Ratio (LCR) across encodings. English modified #5 approaches π to three decimal places.

Visualizations

Preliminary visual inspection suggests the generation process forms a recursive, tree-like branching structure. See the Visualizations page for the full set. Two examples:

Branching tree at level 5 (English modified #1, LCR ≈ 3.037).

Spiral visualization of the recursive generation process.

Algebraic Foundation

Under English modified #5 (LCR ≈ 3.141), the 20 EOA sequences split into two recurrence families. The larger family has 13 sequences that form a true basis for a 13-dimensional solution space of a common linear recurrence. The generating functions share denominator Q₂(x). The other family has denominator Q₁(x) = (x − 1)Q₂(x). The limiting constant ratio is the reciprocal of the smallest-modulus root of Q₂, and it matches π to five decimal places. The 20 sequences satisfy exactly 6 independent linear relations, spanning a 14-dimensional space.

A tool has been developed to ingest each letter's rational generating function and automatically report shared denominators, shared numerators, and numerator-equals-denominator matches across entries. It also computes pairwise GCDs between denominators and flags letters without a match. Initial results are striking and will be published alongside the full algebraic write-up in Part V.

Read the full algebraic foundation →

EOA Program: Research Parts

Part	Title	What It Covers
Part 0	The Engineering of Alphabets as an Open Frontier	Programmatic overview; poses 43 open questions across 10 domains; introduces the EOA program and its scope.
Part I	EOA-43: A Rank-Collapsed Deterministic Word Representation	Presents the 43-term sequence vectors, rank-1 collapse, and the EOA-43 word representation.
Part II	Empirical Cross-Encoding Variation of the Limiting Ratio	Documents LCR variation across 7 encodings and 4 languages; explores encoding-dependence.
Part III	Transient Structure and Detrended Residuals	Analyzes the letter-specific transient regime (terms 1–38) and what information survives.
Part IV	Boundary Mapping on Downstream Tasks	Explores applications in NLP, compression, and adversarial robustness.
Part V	The Operator Beta and Its Recurrence Class	Theoretical classification of the operator; formal characterization of the recurrence and the 13-dimensional basis.

Open Questions

What operator class produces this convergence: substitution system, linear recurrence, contraction mapping, or none?
Why does the 26-letter alphabet collapse to exactly 16 distinct sequence groups?
What is the precise, metric-backed definition of the rank-1 collapse?
Does the generation process exhibit a formal fractal geometry or self-similar branching structure?
Can we design encodings whose LCR approximates π, e, or the golden ratio?
Is there a universal ceiling on LCR, or can an encoding push it arbitrarily high?

Read all 43 open questions →

Publications

The EOA Program, Part 0: The Engineering of Alphabets as an Open Frontier
B. Budargham (2026). PDF · Repository

The EOA Program, Part I: EOA-43: A Rank-Collapsed Deterministic Word Representation
B. Budargham (2026). Repository · arXiv:XXXX.XXXXX

The EOA Program, Part II: Empirical Cross-Encoding Variation of the Limiting Ratio
B. Budargham (2026). In preparation.

The EOA Program, Parts III–V: Transient Structure, Downstream Tasks, Operator Classification
B. Budargham (2026). Future.

Data & Code

EOA Program Repository: Main landing page with overview, parts table, and links to all sub-repositories.
EOA Data: Pre-convergence sequences, figures, notebooks.
EOA Part 0: Paper repository.
EOA Part I: Paper repository.

Contact

For collaboration or access to restricted materials (sequence vectors, oracle, or operator), please contact:
bdarghamneurolabs(at)gmail.com