First, the main text of each independent Piece or Block is analyzed. Artificial intelligence extracts up to 20 important concepts or entities that are genuinely present in the text and have an importance score of at least 0.5. Recognized names are identified through exact vocabulary matching. New names are first searched in the global LeraLink knowledge registry and then, if necessary, in Wikidata. Using the context of the text, the model determines the correct identity among the retrieved options; cases with confidence of at least 0.85 are automatically confirmed, while ambiguous cases remain for human review. In the next stage, the system measures the presence and importance of each concept across the entire knowledge corpus of the same institution: mention frequency relative to text length, presence density, placement in the title, summary, and headings, key definitions, and semantic proximity. The multilingual bge-m3 model, with 1024-dimensional vectors and cosine similarity, is used to measure semantic neighborhood. Connections between Pieces are also derived from shared concepts and entities, vector proximity, references, authorship relationships, and publication structure, and are stored with weights and evidence. Finally, LeraLink produces three levels of knowledge for each concept: a valid general definition, a contextual description based on evidence from the institution's knowledge corpus, and an integrated account. As the knowledge corpus expands significantly, these descriptions are reconstructed while human revisions are preserved. The result is a dynamic multilingual knowledge graph whose semantic connections are traceable and reviewable.

Structural Box A structural box in LeraLink is a statement about the role of several Blocks: this section is a “summary,” a “key definition,” an “exercise,” or a “translator’s note.” The content itself remains within ordinary Blocks, and the role of the box tells it how to be read, displayed, and represented in other media.

Which statement most accurately explains the architecture of inline boxes in LeraLink?

• A box is a separate type of content, and the Blocks within it are removed from the normal Block mechanisms.
• A box is a structural layer that assigns a role to ordinary Blocks, and that same role can be preserved in reading, translation, and print.
• The role of a box is indicated solely by color and disappears in monochrome print.
• Each platform owner can freely define new roles for boxes.

The practical meaning of this approach is that LeraLink can begin with small projects: a ten-piece collection, a booklet, a short course, or a thematic archive. Over time, the same project can become a library, an educational network, a publishing program, or a complete knowledge ecosystem. Product development and project growth are not two separate paths here; every real project helps test the architecture, and each new development increases the capacity of earlier projects.

For example, a researcher can create a project on the history of Iranian political thought and develop each concept, thinker, source, event, or argument as separate pieces. These pieces can first be published as a collection of research notes, then brought together in a long-form article (what I call composition, inspired by music), and later become the basis for an educational course or a book. In more developed versions, the same pieces can be offered in different languages and through different media[1], connected to academic sources, placed in different learning pathways, and recombined[2] according to audience needs.

I will explain the matter of media elsewhere. For now, it is enough to say that every piece of knowledge may have a medium that plays a fundamental role in transmitting it. For example, suppose we turn a poem into an image. The reality is that the primary medium of poetry is “text,” but this does not mean that the poem cannot be presented elsewhere through another medium, such as “image,” “video,” or “audio.” Thus, every independent and coherent unit of knowledge can have one primary medium and several other media.

The recombination of pieces is one of LeraLink’s most significant innovations. Recombination can be one of the most important applications of agentic intelligence, or an agent, in the system. Artificial intelligence, or hooshvareh (هوشواره), can propose new molecules from a collection of pieces, just as in other fields AI has designed, discovered, and can propose entirely new molecules and pharmaceutical compounds. Suppose you have a set of separate, coherent pieces. AI sees shared themes, patterns, and concepts in them and proposes a composition that functions like pharmaceutical formulations, from which wholly different forms of effectiveness can be extracted.

How to get rid of MS Word?