LeraLink Team

LeraLink Team

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.

From Raw Data to a Block of Knowledge: An Example in LeraLink

This Block is an example of how structured data is published in LeraLink. A dataset can be published along with a description, methodological context, source, and calculated variables, and the same data can then be used for a table, chart, or other forms of display. The example below is drawn from a Pew Research Center national survey on Americans’ attitudes toward the country’s two major parties. Respondents were asked whether the Republican Party and the Democratic Party evoke feelings such as frustration, anger, hope, or pride in them.

FeelingRepublican PartyDemocratic PartyDifference in percentage pointsYearPopulationSource
Frustration6475112025U.S. adultsPew Research Center
Anger495012025U.S. adultsPew Research Center
Hope362882025U.S. adultsPew Research Center
Pride2716112025U.S. adultsPew Research Center
The American Trends Panel survey, Wave 180, was conducted from September 22 to 28, 2025. The full survey included 3,445 respondents, and questions concerning these four feelings were asked of a random half-sample of 1,727 people. — Pew Research Center

Trust in Scientists across Selected Countries (An example of LeraLink Data Visualization)

Survey responses are often easier to interpret as proportions rather than isolated numbers. The example below compares the distribution of trust in scientists across four countries using data from the Wellcome Global Monitor.

CountryHigh trustMedium trustLow trustNo opinion
Belgium424882
Spain395461
Australia335881
New Zealand326152
Distribution of the Wellcome Global Monitor Trust in Scientists Index in four selected countries. Percentages represent respondents classified as having high, medium or low trust, or no opinion. — Wellcome Global Monitor 2018, Table 3.2, Trust in Scientists Index.

Body Measurements of Three Antarctic Penguin Species (An example of LeraLink Data Visualization)

Measurements collected from penguins in the Palmer Archipelago, Antarctica, allow several physical characteristics to be compared simultaneously. Plotting flipper length against body mass, for example, makes differences between individual birds and species visible in a way that a table alone does not.

SpeciesBill length (mm)Flipper length (mm)Body mass (g)
Adelie39.11813,750
Adelie39.51863,800
Adelie40.31953,250
Gentoo46.12114,500
Gentoo502305,700
Gentoo48.72104,450
Chinstrap46.51923,500
Chinstrap501963,900
Chinstrap51.31933,650
Selected observations from the Palmer Penguins dataset. Each point represents one measured penguin; species can be distinguished as separate series. — Gorman, K. B., Williams, T. D., & Fraser, W. R. (2014), Ecological Sexual Dimorphism and Environmental Variability within a Community of Antarctic Penguins (Genus Pygoscelis), PLOS ONE 9(3): e90081.

Global Surface Temperature Anomalies, 2021–2024 (An Example of LeraLink Data Visualization)

Temperature anomalies measure how much warmer or cooler a given period is compared with a long-term reference average. NASA’s GISTEMP analysis uses the 1951–1980 period as its baseline. The selected years below offer a compact example of how a time series can reveal change over time.

YearTemperature anomaly (°C)
20210.85
20220.89
20231.17
20241.28
Annual global surface temperature anomaly relative to NASA’s 1951–1980 baseline. Values shown here are a selected recent subset of the GISTEMP record. — NASA Goddard Institute for Space Studies, GISTEMP v4 — GISS Surface Temperature Analysis. GISTEMP is continuously updated as new observations and corrections become available. NASA GISS Data

An Institution is an independent institutional page within LeraLink. It represents a distinct part of an organization that needs its own public identity and its own way of presenting knowledge. For example, one organization may create separate Institutions for its research center, graduate school, publication unit, public education platform, or community initiative. Each can have its own landing page, content, and knowledge structure while still belonging to the same larger organization. Institutions are modular. They can be created and shaped according to the structure and needs of each organization. An Institution is therefore not necessarily a separate legal organization. It is the public-facing space through which an organizational unit presents its identity, activities, and knowledge in LeraLink.

LeraLink organizes knowledge through a set of connected levels, from public institutional spaces down to small reusable units of knowledge. The structure is: Institution → Program → Dossier → Series → Volume → Composition → Piece → Block An Institution is the top-level public-facing space in this structure. An organization can create more than one Institution for different parts of its work, such as a research center, school, publication, community, or educational initiative. Within an Institution, knowledge can be organized into Programs, Dossiers, Series, Volumes, and Compositions. A Composition brings Pieces together into a meaningful work, such as an article, chapter, lesson, or similar structure. A Piece is the smallest meaningful and reusable unit of knowledge. Each Piece is made of one or more Blocks, such as text, images, quotations, audio, video, or other content elements. This structure allows organizations to present complex bodies of knowledge while keeping individual knowledge units small, reusable, and connected.

An Entity is a clearly identified person, organization, place, concept, event, work, or other subject connected to knowledge in LeraLink. Entities help LeraLink understand what a Piece or Source is about. For example, a Piece about Paulo Freire may be connected to the Entity Paulo Freire, the concept Critical Pedagogy, and the work Pedagogy of the Oppressed. Entities can be connected to many Pieces and Sources. This creates a shared network of knowledge and helps users discover related material across different projects, languages, and formats. Unlike a keyword or tag, an Entity has its own identity and can carry structured information and relationships.

A Source is the original material that supports, informs, or provides evidence for a Piece or a Block. A Source can be a book, article, research paper, chapter, report, website, lecture, podcast, video, dataset, archive, or another type of reference. Pieces can be linked to one or more Sources. This helps show where the knowledge comes from and makes it easier to verify, explore, and reuse responsibly. A Source is not the same as a Piece. The Source is the reference material; the Piece is the smaller unit of knowledge created, extracted, summarized, explained, or connected from it. Linking Pieces to Sources helps keep knowledge traceable and transparent.

A Block is the basic content element used inside a Piece. A Block can be a paragraph, heading, quotation, image, audio clip, video, list, embed, or another content type. Several Blocks can be combined to form one Piece. A Piece represents a meaningful unit of knowledge. A Block represents how that knowledge is structured and presented. For example, one Piece may contain a short explanation, an image, and a quotation. These are three separate Blocks inside the same Piece. Blocks make it easier to edit, arrange, reuse, and display content in different formats while keeping the Piece as one coherent unit.

A Piece is the smallest meaningful unit of knowledge in LeraLink. A Piece can contain an idea, explanation, argument, example, quotation, image, audio, video, or another form of useful knowledge. It should be small enough to reuse, but complete enough to make sense in context. Each Piece has its own identity. It can be connected to sources, translated into other languages, reviewed, updated, linked to other Pieces, and placed in different structures. A Piece does not belong permanently to one article, lesson, or course. The same Piece can be reused in different contexts without being recreated. This makes knowledge in LeraLink modular, traceable, flexible, and easier to maintain.

A Content Management System (CMS) is mainly designed to create, organize, and publish content such as web pages, articles, posts, or media. LeraLink is designed around knowledge rather than pages. In a typical CMS, content is usually created for a specific page or publication. In LeraLink, knowledge is stored as smaller, independent pieces that can be connected, translated, reviewed, reused, and placed in different structures without being copied. A CMS helps you manage content. LeraLink helps you manage the relationships, meaning, reuse, and development of knowledge. This makes it possible to use the same knowledge in different articles, guides, courses, collections, or learning experiences while keeping its identity and history intact.

A Learning Management System (LMS) is mainly designed to deliver and manage courses. It usually organizes learning around fixed structures such as courses, modules, lessons, assignments, and enrolled learners. LeraLink starts from a different point: the knowledge itself. In LeraLink, knowledge is created as small, reusable pieces that can exist independently of any single course. The same piece can be translated, reviewed, connected to other knowledge, reused in different learning paths, or included in several projects without being duplicated. An LMS usually asks, “How do we deliver this course?” LeraLink also asks, “How can this knowledge remain useful, connected, reusable, and accessible across different languages and contexts?” LeraLink can support courses and structured learning, but the course is only one possible way of organizing the knowledge.

LeraLink is a multilingual knowledge and learning platform built around small, meaningful units of knowledge. Instead of treating knowledge as fixed inside books, courses, articles, or videos, LeraLink breaks it into smaller pieces that can be connected, translated, reviewed, reused, and combined in different ways (For instance: This very answer is a Block of knowledge).  These pieces can include ideas, explanations, arguments, examples, quotations, images, audio, video, and other forms of knowledge. LeraLink makes it possible to organize these pieces into larger structures such as articles, learning paths, guides, lessons, and courses. The same knowledge can also appear in different languages, formats, and contexts without being recreated from scratch. The goal is to make high-quality knowledge easier to access, understand, connect, contribute to, and share.