Banco de Estructuras
de Lenguaje Accesible.
Bank of Accessible Language Structures. A systematic, traceable method for producing documents people can actually understand: plain-language and easy-to-read versions of a ruling, a medical report, a notice.
In the original, what the reader needs comes last. In the accessible version, each block answers one of their questions, in their order. Example from the BELA-Madrid Comprensible pilot.
Today, text accessibility is not systematically auditable
Current plain-language adaptations depend on each adaptor’s individual judgment: two professionals produce two different versions of the same text, and neither can be audited. A procedure is missing.
The cause is conceptual. Corrective practice — style guides, training, automated assistance — almost always intervenes at the level of the sentence and the word. Yet only a small part of the recommendations in the international plain-language standard itself concerns vocabulary; the rest concerns structure and design. It is precisely that structural level that we have least learned to operationalize.
Readers don’t process isolated sentences: they build a global representation of the text. That is the level at which accessibility has to intervene.
From correcting sentences to formalizing architecture
BELA™ identifies the discursive macrofunctions that organize each document genre: the blocks that answer the reader’s questions — what am I being told, what was decided, on what grounds, what can I do now — and diagnoses whether each one is present, locatable, and understandable.
Its core is a bank of previously validated text structures: instead of rewriting each document from scratch, the adaptor instantiates a structure already proven for that type of text. The content changes; the architecture that guarantees comprehension stays the same.
Macrofunctions are not a fixed catalogue: they are induced from the corpus of each genre. Once identified, they are cross-referenced with the four principles of plain language in a matrix that makes every step of the adaptation explicit and auditable. The example shows the four macrofunctions of the legal-administrative pilot.
| ISO ↓ · Macrofunction → | EXPLANATION | DECISION | REASONS | GUIDANCE |
|---|---|---|---|---|
| Relevance | · | · | · | · |
| Findability | · | · | · | · |
| Understandability | · | · | · | · |
| Usability | · | · | · | · |
Every cell is a verifiable operational criterion: traceability stops being a promise and becomes a procedure. In the clinical pilot the columns are six different ones; the matrix is the same.
Each text is annotated in three layers:
Fidelity above all: the method never invents content that is absent from the original. If the text leaves something unresolved, the accessible version communicates that uncertainty just as faithfully.
Two very different genres, one method
The current pilots are a starting point, not a boundary: the framework is designed to transfer to any kind of document a person receives and needs to understand.
BELA-Madrid Comprensible
49 public-information-access rulings from Madrid City Council (2018–2026), released as open data through its Transparency Portal.
Four macrofunctions organize the diagnosis:
BELA-Consulta Comprensible
Primary-care consultation reports (MEDICLARO corpus), in collaboration with the HULAT group at Universidad Carlos III de Madrid.
Six macrofunctions, induced from the corpus itself:
A formal process for every new genre, corpus, or language
A bank built for administrative rulings doesn’t simply carry over to consumer contracts or drug leaflets. Each new domain requires the same process:
Corpus
Gather a sample of documents that is sufficient and representative of the target genre and language.
Criteria
Review the diagnostic criteria to incorporate the linguistic and institutional specifics involved.
Bank
Build a new bank of structures from that corpus.
Vocabulary
Adapt the controlled lexical references to the corresponding institutional terminology.
Validation
Validate the easy-read versions with real users according to the applicable standard.
The bank is not the final product: it is the evidence that the method works
Which is why the work doesn’t end when a bank is complete. What transfers is not a repertoire of templates but a diagnostic procedure that can be applied again whenever it is needed.
And it is needed often: regulation changes, and a compliant structure stops being compliant; organizations don’t need templates, they need auditing — knowing what fails in their own documents and being able to show they have fixed it; and every new sector is a new bank, because each genre organizes what the reader needs to know in a different way.
Artificial intelligence, as a governed process
Several recent studies point in the same direction: language models simplify vocabulary and syntax, but rarely analyze a document’s architecture unless someone has formalized it first. BELA™ formalizes precisely that architecture.
AI intervenes at two moments — analytical in the diagnosis, generative in the drafting — but never before the architecture has been formalized, always within structures fixed by the method, and no version enters the bank without expert approval.
Video · BELA demo
The method at work on a real document: from the original text to the structural diagnosis and the accessible version.