Golden Piece: The Foundation of Knowledge Architecture in LeraLink

In What Unit Does Knowledge Endure?

A note may be written while reading a book, revised in a group discussion, incorporated into part of an article, and later used in a lesson in another language. With each move, questions remain: what was the note about, what evidence did it rely on, who changed it, and which version is the reader seeing now? I design LeraLink around the continuity of these very relationships. The matter is preserving something that can be revisited, disputed, and put to use in a new context, without losing its history and responsibility.

The architecture of knowledge shapes the relationships among people, language, tools, and ways of working; a relationship that was also central to Engelbart’s augmentation of human intellect. Nelson’s idea of transclusion sought to preserve the link between the text used and its origin; nanopublication likewise links an assertion, its provenance, and publication information in a small package. LeraLink’s proposed contribution is to formulate an architecture that connects these concerns across the cycle of writing, translation, collaboration, peer review, and publishing, and provides a defined place for text, image, sound, and data.

In this framework, a “unit of knowledge” is the name for a matter that can be addressed. Recording a claim does not make it true, just as the value of a poem, narrative, or musical performance cannot be assessed by the criterion of propositional truth. A Piece must keep clear the kind of work it is, the context for understanding it, and the appropriate way to evaluate it. “Completeness,” too, is assessed in relation to function: the reader should know what they are encountering and where to turn to understand or examine it. No Piece is independent of all languages, experiences, and prior knowledge. We can call this unit of knowledge an “episteme.”

Block, Piece, and the Ontology of Knowledge

Block: The Unit of Writing, Recording, and Reuse

A Block^[1] is the unit with which we write, record, and compose: a paragraph, quotation, captioned image, recorded sound, or dataset can be a Block. A Block within a Piece has a structural identity and a relationship to versions and translations; an independent Block^[2] can also remain a draft, be kept private, or be published by decision of its rights holder, and acquire an address, a translation, and the possibility of appearing in other works. Thus, having an address or the capacity for publication alone does not determine the boundary between a Block and a Piece. To record an idea, an author need not turn it into a complete work from the outset in order to give it the opportunity for public life.

A media asset, such as an image file or recorded sound, is the material infrastructure of content. A Block gives that asset a role and context: it makes clear what the image is evidence of or to which statement the sound belongs. A Block may also contain text, image, and sound at the same time. Thus, being a Block, media type, and work type are three distinct attributes. Recognizing this difference matters for the gradual growth of knowledge: raw materials can be gathered before the final design becomes clear, without mistakenly being presented as approved works.

Piece: the unit of coherence and authorship responsibility

A Piece is the smallest sufficient composition for a specific knowledge function, with an identifiable identity, versioned body, linguistic representations, provenance, rights, and authorship responsibility. The body of a Piece is made of one or more Blocks and may refer to existing Blocks or Pieces. The transition from Block to Piece becomes meaningful when someone accepts responsibility for its selection, arrangement, contextualization, and presentation as a whole. The scale of this whole may vary from a contextualized definition to an essay; word count or page length is not the determining criterion.

Composition in LeraLink can be carried out by referring to the identity and revision of a Block. The presence of a published Block in another work is not the same as transferring unpublished material to a new place of authorship; likewise, reuse of another person's writing does not grant the user the authority to edit its original. The authorship value of a Piece also lies in the relationship established among its components (Blocks): two compositions based on the same evidence may construct two different arguments. Preserving the provenance of components must be accompanied by recording the responsibility of the composer (the author of the Piece).

The diagram of the stages of producing a Piece and placing it within other structures shows one of the pathways through which knowledge is formed and reused. Movement within the LeraLink architecture can also continue in the direction of extraction and revision: from studying a Piece, a new Block may emerge, and from examining that Block, a new composition may be created. Extraction must retain an indication of the previous context, and the new composition must also be recognized as an authorship decision.

Structures: identity, placement, and governance

The LeraLink organizational ontology separates the type of work from its place within a structure. A Composition can place several Pieces alongside one another in a continuous reading; Volume, Series, and Dossier are also forms for organizing works. These names do not constitute a mandatory ladder, and not every work needs to pass through all levels. The distinction among work, expression, and manifestation in the IFLA Library Reference Model also provides a basis for thinking about the difference between the identity of a work and its forms of presentation; the LeraLink model is not a one-to-one mapping of that model.

Placement^[3] is a relationship that defines the presence of a Piece in a specific context, with a defined order, role, language, and mode of display. A Piece may be present in multiple structures without each presence counting as a new work. A Program is the space for managing work with specific goals and policies; the environment of an individual or Institution likewise defines the boundaries of membership, access, and resources. A diagram of concepts and entities is another layer for discovering relationships. The reading structure, the network of meaning, and the boundaries of authority must be able to work together without taking each other's place.

Constitution of the Golden Piece

Principle One: Identity Must Remain Traceable Across Different Forms

Every Piece must have a clear reference identity that is not lost when its address, display language, or location of presence changes. The reading page, wall card, quotation, and generated file are forms of presentation. By contrast, an adaptation or independent publication by another publisher may create an entity with a new identity and responsibility. In such cases, the relationship of derivation or publication must be recorded. The principle of a single identity does not authorize treating as one versions with different creators, rights, or peer review status.

Principle Two: Citation Must Reach a Determinate State of the Work

The citable body must be specified by language and publication version. A footnote, data, accepted translation, and any reference that plays a role in an argument must be read in relation to that same version. «Orbit»^[4] is the place where explanations, conversations, relationships, and surrounding reports grow; changes to Orbit must not alter the cited content without notice. A substantive correction must have a clear history and, if the work is withdrawn, an understandable record of its status must remain to the extent permitted by rights and privacy.

Principle Three: Composition Must Preserve Context and Version Policy

Each reused component must preserve its origin and its relationship to the host work. Pinning to a specific version and tracking the latest version are two different policies and must be clear to the work owner. In a citable publication, the versions of components that affect meaning must also be specified; keeping the host text unchanged alone is not sufficient for the stability of the work as a whole. This requirement is the stricter standard of this revision for completing the guarantees of composition, not a claim that all dynamic dependencies can now be fully pinned.

Principle Four: Language and Medium Must Not Create a Hierarchy of Authority

Every language and every carrier of meaning must, in accordance with its function, permit identification, attribution, and evaluation. The language in which composition began and the translation relationship must be recorded, because they are important to understanding the work; chronological precedence does not grant the source language epistemic privilege. Image and sound may also be the primary content. Explanatory text, transcription, translation, and summary have different relationships to the work and should not be presented as entirely equivalent forms, yet they should not be separable from the work.

Principle Five: Provenance Must Be Auditable

The provenance record^[5] must make clear who or which system, through what process, produced or altered what, and which decision accepted it. The entity, activity, and agent model in the Web Provenance standard is a useful reference for this distinction. The source entity's being unknown must be made explicit. A content fingerprint can reveal change, but does not by itself establish the validity of a claim, the creator's consent, or the credibility of peer review; each requires its own appropriate evidence.

Principle Six: Automation Must Respect the Boundary of Human Responsibility

An intelligent service must have a limited task, known provenance, and the possibility of correction. In LeraLink's current artificial intelligence architecture, the writing of and responsibility for the Piece belong to a human; a translation or composition suggestion without a human decision does not become approved content. Rather, the boundary between machine and human actions is made clear. Indexing and part of the construction of the concept layer may be automated, because they are derived from the text and do not rewrite the author's text. These outputs must remain identifiable and correctable, and must not erase human corrections upon re-execution^[6].

Principle Seven: Participation, Authority, and Ownership Must Be Recorded Separately

Editing rights, editorial responsibility, participation credit, and ownership of the work are four independent relationships^[7]. Translators, editors, data researchers, designers, and speakers must be recognized and made visible in proportion to their work; an invitation to collaborate must not by itself make someone an author or owner. The classification of participation roles provides an important precedent for making visible the work behind the work and the people behind it. LeraLink extends this concern to multilingual and multimedia work. Any future participatory economic model also requires a separate agreement; recording a share in production does not mean automatically assigning a financial share, and this matter falls within the bounds of human judgment.

Principle Eight: Credibility Must Depend on Evidence and the Scope of Peer Review

An evaluation must make clear what was examined, in which version, by what method, and according to which criteria. Views, likes, and semantic proximity are signs of circulation or connection, not proof of a work's quality. In LeraLink's publication and peer review model, the validity of peer review is tied to a specific publication and to its process evidence. A translation, subsequent revision, or different version by the author must not automatically inherit that validity. The evaluation method must also be appropriate to the type of work.

Ninth Principle: Access Must Remain Valid Throughout the Circulation of a Work

Authorization must be observed in reading, searching, embedding, file retrieval, and collaboration. Being private is not the same as being a draft, and being published is not the same as being public. An organizational role is valid only within the scope of delegated authority. A reader of a publication must not gain access to confidential materials through related links or files. Transparency of publication history must also be compatible with the confidentiality of peer review and personal data; accountability does not require disclosure of every detail.

Tenth Principle: Gradual Growth and Exportability Must Remain Possible

Initial registration must involve little friction, and the addition of structure, source entity, translation, and evaluation must be able to proceed gradually. At the same time, the user must be able to export their content from the system together with the information needed to understand and reuse it. The principles of findability, accessibility, interoperability, and reusability support this commitment. Exportability does not end with obtaining an image or readable file; preserving relationships and rights also matters. Accessibility may also require authentication and compliance with authorization. Therefore, LeraLink must be able to provide the work's owner with a complete package of all these items while preserving privacy boundaries.

Multilingualism Within the Structure Itself

Independence from language is a property of infrastructure; meaning cannot be separated from language and history. LeraLink keeps the user interface language separate from the language of knowledge, so that, for example, a Persian environment can host a Dutch text and an Arabic Block without the language of the tool determining the language of the work. Linguistic representations are connected to a shared identity, but they have their own author, review status, and relationship to the source version. Human translation must be protected from automatic rewriting, and a change in the source must make the need for review apparent; a language that has not yet been updated must not, without explanation, appear to be the work's latest expression.

The data Block is a clear example of structural multilingualism: in the LeraLink data model, values and graph structure are shared across languages, while labels and descriptions are language-specific. At the concept layer as well, separating identity from name, like the distinction between concept and label in knowledge organization standards, enables multiple names and linguistic forms. However, similarity of terms should not obscure differences among intellectual traditions. The architectural response to epistemic injustice is to provide the possibility of presence, attribution, and contestation; the elimination of inequality cannot be inferred merely from the existence of translation. The dominant and reductive understanding of linguistic-epistemic injustice, which in practice governs the Europe-centered idea of multilingualism, is one of the issues challenged by LeraLink’s new architecture. Considerable costs are spent on this dominant understanding and, as a result, despite substantial investment, it has had a very superficial impact on the category of multilingualism across the European technological landscape and even in the open access movement.

The network of meaning, the context of reading, and learning

The LeraLink knowledge layer distinguishes between the identity of a concept or entity^[8], evidence of its presence in works, its introduction in the context of a dossier, and the Piece about which the author has written. An encyclopedic definition and an account of a concept’s place in an author’s writings do not perform the same task. Linking a name to Wikidata is likewise merely an aid to identifying the referent. Similarity, co-presence, and reference must be recognized as distinct relations; the juxtaposition of two concepts cannot be used to infer mutual confirmation or a causal relationship.

In this architecture, learning can begin with reading itself. Explaining a term, comparing two pieces of evidence, annotation, and self-guided questioning are ways of engaging with a work. Educational and editorial containers distinguish the role of an example, definition, or editor’s note from mere visual arrangement; the Web Annotation Model is also a reference for preserving the relationship between a note and its subject. Linking a Block and a Piece makes it possible for the result of reading to become material for a new composition. The educational and pedagogical effects of these capabilities, such as reducing cognitive load or improving learning, still require empirical research and further development.

Collaboration and publication, without losing roles and responsibility

Collaboration at the Block level allows several people to work on a Piece while keeping the boundaries of their work clear: the author can advance the text, the translator can refine its expression in another language, and the data specialist can work on the Block related to data. The shared draft and version history have a single reference point; the authority to alter the text is not the same as the authority to finalize a version and publish it. The desired outcome is collaboration in which simultaneous work does not lead to ambiguity in attribution or the removal of responsibility for decisions.

In institutional publishing, the venue that accepts a work^[9] ; may have a defined policy and editorial group (editorial board). A reviewer must examine a specified package of the work, and the policy governing that round of peer review must also remain in the record. An author’s version and a publisher’s publication can be linked to one another and have independent histories. The LeraLink press also produces digital or print output in a separate layer so that layout decisions do not alter the reference text. A complete preservation package and a reading-friendly file serve two different purposes.

The distinction of LeraLink and the boundary of claims

Content reuse has a history: synchronized Blocks in Notion and structured authoring in the DITA standard^[10] have previously addressed this need; Open Journal Systems also supports editorial workflows, peer review, and multilingual publishing, but not at the Block level. LeraLink's proposed point of distinction is the linking of an independent Block, authorship responsibility, language representations, knowledge relationships, and versioned publication in a single working model. The quality of this linkage must be assessed through real-world scenarios and error testing; merely placing capabilities side by side does not prove superiority over specialized tools, but LeraLink has taken appropriate steps toward creating an editor suited to this approach and is in the stage of testing, gathering feedback, and improving it.

The application of the model can be seen in different projects: a historian uses notes connected to a document in an essay; a music instructor brings together performance audio, an image of the score, and interpretation in an educational Piece; a research group reuses data and methodological explanation across several reports; a writer, too, brings scattered notes into a coherent structure. The shared core is a clear relationship between part and whole, preservation of provenance, and the possibility of revision. The specialized needs of each field should be developed along this core, without reducing all knowledge activities to the model of the academic article (which, unfortunately, despite the remarkable growth of technology and innovative architectural possibilities, is still the “status quo”).

In the version reviewed, the independent Block, referential composition, translation management, collaboration, concept layer, editorial workflow, file production, and structured output package are present in the code and documentation; access to them depends on the permissions of each environment. Exchange between environments on the same platform is not equivalent to operational communication between independent systems. Broader interoperability, with reference to the linked data format and the packaging of research objects, is the horizon for development and assessing compatibility. LeraLink has not yet reached this stage and does not claim that it can attest to full compliance with those standards or the realization of the federated network proposed in the white paper, but it has taken effective and decisive steps along this path.

What Makes a Piece “Golden”?

A golden Piece is a reference example for testing the fidelity of the architecture to its commitments. Being golden does not mean conferring a seal of truth, popularity, or accumulating every capability in one place. The reference example must be able to show how translation, revision, collaboration, reuse, and access restrictions take place without compromising identity, attribution, and history. Supplementary criteria must be proportionate to the type of work: poetry does not require the same peer-review dossier as a research article, and a Piece without images is not incomplete simply because it has no images. Any claim of compliance must have a verifiable scope, version, and test.

LeraLink’s fundamental commitment is that, as knowledge grows and circulates, the possibility of attending to it should not be lost. A reader should be able to trace the path of a statement back to its context and source entity; an author should be able to recognize their contribution and correct their error; collaborators and publishers, too, should remain accountable within the limits of their authority. The durability of knowledge depends on such possibilities: that works can be read and created anew in other languages, generations, and institutions, while their relationship with those who played a role in their creation remains identifiable.