Published November 1, 2026 | Version v1

Marsji: A Concept-ID Semantic Layer for Cross-Language Lexical Mapping and Accessible Interfaces

Description

# Marsji: A Concept-ID Semantic Layer for Cross-Language Lexical Mapping and Accessible Interfaces

 

**A Proposal and Status Report**

 

**Author:** Gino Rouzee

**Affiliation:** Independent researcher

**First version:** May 13, 2026 (Marsji Day)

**This version:** November 8, 2026

**Repository:** GitHub/Marsji-Language-Project

**Web interface:** marsji.com

**DOI:** [10.5281/zenodo.23857698](https://doi.org/10.5281/zenodo.23857698)

 

---

 

## Abstract

 

This paper describes Marsji, a semantic layer that assigns a fixed identifier (ID) to each concept and links that ID to words in many natural languages. Instead of mapping words directly from one language to another, every word is mapped to a shared concept ID, and from there to its counterparts in other languages. The current dataset contains approximately 2.084 million word-to-concept relationships across 50 languages, including a core vocabulary of roughly 1,000 structural words per language.

 

The paper presents the three-layer architecture of the system, the design of its ID scheme, the current state of the dataset including its known quality limitations, and what the approach can and cannot do. It also proposes, as a direction for future research, a learnable tactile and visual code for concept IDs, inspired by the six-dot braille cell, with a variant that can be formed with one hand, intended as an additional communication aid for blind, deaf and deafblind users. The claims in this paper are limited to what the current implementation demonstrates; the accessibility application is presented as a hypothesis to be tested together with the communities concerned.

 

## 1. Introduction

 

Most translation technology today relies on statistical or neural models that learn correspondences between languages from large amounts of text. These systems are powerful, but their output is not deterministic: the same input can produce different translations, and it is difficult to trace why a particular word was chosen.

 

Marsji explores a complementary approach. Its starting point is the observation that many concepts, such as *love*, *water* or *mother*, are shared across languages even though the words for them differ. If each such concept receives a stable ID, then words in different languages can be linked through that ID in a transparent and reproducible way. A lookup from Urdu to Spanish, for example, becomes: Urdu word → concept ID → Spanish word.

 

The guiding idea behind the project can be summarized as: *the power of information lies not in the number of symbols, but in the relationships between them.* Marsji is an attempt to make those relationships explicit, stable and machine-readable.

 

Marsji is not intended to replace natural languages, nor to replace machine translation. It is a structured lexical layer that can support search, indexing, vocabulary learning and, potentially, accessibility tools.

 

## 2. Related Work

 

The idea of a shared conceptual layer between languages has a long history, and Marsji builds on several existing approaches:

 

- **Universal Networking Language (UNL)**, developed under the United Nations University from the late 1990s, used "Universal Words" as language-independent concept identifiers, with the aim of serving as an interlingua for translation.

- **WordNet** (Miller, 1995) organizes English words into sets of synonyms representing distinct senses, linked by semantic relations. Many WordNets for other languages have since been linked to it.

- **BabelNet** (Navigli & Ponzetto, 2012) combines WordNet and Wikipedia into a large multilingual semantic network in which concepts carry identifiers across hundreds of languages.

- **Wikidata lexemes** link words in many languages to language-independent items.

- **Blissymbolics** (Bliss, 1965) is a system of symbols representing concepts rather than sounds, used today in augmentative and alternative communication.

 

Marsji differs from these mainly in three respects: it introduces its own compact word forms for each concept (the Marsji core vocabulary), it adds a deterministic reversible encoding layer, and it explores a direct link between concept IDs and a tactile/visual code that can also be formed with one hand. A systematic comparison with BabelNet in particular is part of the planned evaluation (Section 8).

 

## 3. System Architecture

 

Marsji consists of three layers.

 

### 3.1 Surface layer (natural language)

 

The words of natural languages as they are written: *love* (English), *amour* (French), *محبت* (Urdu). Words with several meanings (homonyms and polysemous words) are stored as separate entries, one per sense, so that each entry can be linked to exactly one concept.

 

### 3.2 Matrix layer (reversible encoding)

 

Each surface entry is converted into a deterministic machine string using a reversible transformation: the original word can always be reconstructed exactly from its encoded form. For example, the 45-letter word *pneumonoultramicroscopicsilicovolcanoconiosis* maps to a 45-character encoded string and back without loss.

 

This layer provides a uniform, language-independent representation for processing. It does not compress data; its purpose is consistency and exact reversibility.

 

### 3.3 Semantic layer (Marsji core)

 

Each concept receives a fixed ID and a short Marsji word form, for example *nowe* for the concept of love. All surface entries that express this concept, in any language, are linked to the same ID.

 

## 4. The ID Scheme

 

Every concept ID must be unique and stable over time, so that links between languages never break when the dataset grows.

 

In the current implementation, each concept receives a sequential numeric ID. Once assigned, an ID is never changed or reused, even if the concept is later refined or an entry is removed. New concepts always receive new IDs. This guarantees that any reference to an ID, in the dataset or in external systems, keeps pointing to the same concept.

 

The hard problem in any concept-based system is not numbering but **disambiguation**: deciding which concept a word refers to in a given context. The English word *bank* can refer to a financial institution or the side of a river. Marsji stores these as separate entries with separate IDs, but choosing the correct one in a running text still requires context, either from the user or from an additional disambiguation step.

 

## 5. Current Dataset

 

| Item | Status |

|---|---|

| Word-to-concept relationships | approx. 2.084 million |

| Languages | 50 |

| Core structural vocabulary | approx. 1,000 words per language |

| Specialized vocabulary | medical nomenclature (partial) |

| Formats | JSON and CSV tables, expansion scripts |

| Platforms | Linux, macOS, Windows (no external dependencies) |

 

**Data quality.** A large part of the dataset was generated with the assistance of AI language models and has not yet been fully validated by native speakers. During development, several types of errors were identified, including invented abbreviations that conflict with established ones (notably in medical terminology) and outdated information presented as fact. Validation of the core vocabulary by native speakers is a priority. Users of the dataset should treat entries outside the validated core as provisional.

 

## 6. Capabilities and Limitations

 

**What Marsji can do today:**

 

- Look up a word in one language and return its counterparts in other languages via a shared concept ID

- Do so deterministically: the same input always gives the same output, and every link can be inspected

- Provide stable concept IDs that can be used for multilingual search and indexing

- Run locally without machine-learning models at lookup time

 

**What Marsji cannot do (yet):**

 

- Translate full sentences. Sentences involve grammar, word order, tense, agreement, politeness forms and idioms that cannot be handled by word-to-concept mapping alone. *"It's raining cats and dogs"* contains no word that refers to heavy rain.

- Disambiguate automatically. Selecting the correct sense of a word in context requires additional information or processing.

- Guarantee the correctness of entries that have not yet been validated (see Section 5).

 

## 7. Proposed Application: A Tactile and Visual Code for Concept IDs

 

*This section describes a research direction, not a validated system.*

 

### 7.1 Idea

 

The braille cell uses six positions arranged in two columns of three:

 

```

(1) o   o (4)

(2) o   o (5)

(3) o   o (6)

```

 

The proposal is to define a code in which Marsji concept IDs are expressed as sequences of such six-position patterns. The same code could be:

 

- **felt**, through finger positions or a flat haptic plate with micro-actuators, for blind and deafblind users;

- **seen**, as the corresponding hand shapes and movements, for deaf users.

 

Because each pattern refers to a concept rather than to a word in a particular language, two people who have both learned the code could, in principle, exchange basic meanings regardless of their spoken language. A message entered as a sequence of concept codes could then be rendered in the receiver's own language through the concept ID.

 

### 7.2 Design choices

 

The code was developed through a series of deliberate design decisions:

 

- **Origin in finger counting.** The starting point was finger-counting methods such as Chisanbop, in which the fingers of both hands represent numbers. These methods show how many distinct combinations a hand can express.

- **Avoiding numerical association.** In finger counting, one hand represents units and the other tens. When the same positions are used for concepts, people tend to read them as numbers and start counting. The Marsji code therefore deliberately reorganizes the positions, so that a pattern is perceived as a concept rather than a quantity.

- **A two-handed form and a one-handed variant.** In the standard form, the two columns of the cell are formed with two hands. Many signs in established sign languages, such as American Sign Language (ASL) and British Sign Language (BSL), require two hands, which disadvantages people who have the use of only one hand. The Marsji code therefore also has a one-handed variant, so that nobody is excluded.

- **Palm orientation in the one-handed variant.** To express all six positions with one hand, positions 1, 2 and 3 are formed with the palm facing forward, and positions 4, 5 and 6 with the palm facing backward. This rule applies only to the one-handed variant. For a sighted deaf receiver, the orientation is visible. For a blind receiver reading the hand by touch, the palm (soft and smooth) and the back of the hand (knuckles, tendons, nails) feel distinctly different, which is expected to make the orientation recognizable by touch. How quickly and reliably this works in a running conversation remains to be tested.

 

### 7.3 Example scenario: a broadcast from Pakistan watched in Mexico

 

A news anchor in Pakistan speaks Urdu. An interpreter renders the broadcast not in Pakistani Sign Language, but in the Marsji concept code: each concept is shown as its fixed hand pattern, linked to its concept ID. A deaf viewer in Mexico who has learned the same code watches the interpreter. Because the viewer associates each pattern with the same concept ID, and that ID with the Spanish word, the viewer reads the message directly in Spanish, without knowing Urdu and without translation software.

 

The principle is the same as with numerals or traffic signs. The symbol *7* is read as *sieben*, *sept* or *siete* depending on the reader, yet everyone understands the same quantity. The Marsji code applies this principle to concepts.

 

Conditions and limits of this scenario:

 

- Both the interpreter and the viewer must have learned the Marsji code.

- The interpreter must convert spoken Urdu into concept codes in real time, which is a skill that requires training.

- The viewer receives a sequence of concepts, not a grammatically complete Spanish sentence. Word order follows the interpreter's choices (Urdu typically places the verb at the end, Spanish usually does not), and tense, number and other grammatical details are only conveyed to the extent that the code contains concepts or markers for them. The core meaning comes across; the full nuance of the original may not.

 

### 7.4 Important clarifications

 

- **This would be a new, learned code.** Users would need to learn it, just as braille or any other code must be learned. Nobody would understand it automatically.

- **It does not replace sign languages.** Sign languages, such as Pakistani, Mexican or Flemish Sign Language, are complete natural languages with their own grammar and vocabulary. They are not equivalent to braille, which is a writing system for spoken languages. The proposed code would at most be an additional bridge for specific situations, such as direct communication between a blind and a deaf person.

- **It would share the limitations of Section 6.** A sequence of concepts conveys basic meaning but not the full grammar and nuance of a language.

 

### 7.5 Development approach

 

Any work on this application should be carried out together with blind, deaf and deafblind people and their organizations, who are the experts on their own communication needs. The first step is a small learnability study.

 

## 8. Evaluation Plan

 

1. **Accuracy of the core vocabulary:** native speakers review a random sample of entries per language; error rates are reported per language.

2. **Coverage comparison:** compare concept coverage and alignment with BabelNet for a set of common concepts.

3. **Disambiguation:** measure how often words in real sentences map to more than one concept, and test methods to select the correct one.

4. **Learnability of the tactile/visual code:** measure how long participants need to learn a first set of, for example, 100 concept codes, and how accurately they recognize them afterwards, both visually and by touch.

 

Results will be published in the repository, including negative results.

 

## 9. Limitations

 

Beyond the points in Sections 5 and 6: the project has so far been developed by a single independent researcher, without external peer review. The dataset's size has grown faster than its validation. The accessibility application has not yet been tested systematically. These limitations are stated openly so that others can assess, reuse and improve the work.

 

## 10. Conclusion

 

Marsji is a working multilingual lexical layer in which words from 50 languages are linked through stable concept IDs, offering deterministic and inspectable cross-language lookup. It is not a complete translation system, and its larger ambitions, in particular a concept-based tactile and visual code for accessibility with a one-handed variant, remain to be tested. The next steps are validation of the core data, systematic comparison with existing resources, and careful, community-led exploration of the accessibility idea.

 

## References

 

- Bliss, C. K. (1965). *Semantography (Blissymbolics)* (2nd ed.). Sydney: Semantography Publications.

- Miller, G. A. (1995). WordNet: A lexical database for English. *Communications of the ACM*, 38(11), 39–41.

- Navigli, R., & Ponzetto, S. P. (2012). BabelNet: The automatic construction, evaluation and application of a wide-coverage multilingual semantic network. *Artificial Intelligence*, 193, 217–250.

- Uchida, H., Zhu, M., & Della Senta, T. (1999). *The UNL, a Gift for a Millennium*. Tokyo: United Nations University, Institute of Advanced Studies.

- Rouzee, G. (2026). *Marsji — A Semantic Language Layer*. Zenodo. https://doi.org/10.5281/zenodo.23857698

- Marsji Language Project. GitHub repository: GitHub/Marsji-Language-Project.

Files

Marsji_paper_revised.md

Files (13.8 kB)

Name Size
md5:dbacec5486f816fd04cfc47e9a68ef04
13.8 kB Preview Download