Educational AI Daily translators' guide
How the paper is translated from Spanish into Catalan, Galician, Basque and English. It is read by every agent that translates something the public will read, together with the style book, which applies equally in every language. Each assignment uses the general section and the section for its own language.
Automatic AI translation of the Spanish original.
For all languages
- The meaning is translated, in the natural register of each language, as a journalist writing in that language would put it: no calques from Spanish in word order, prepositions or expressions.
- The same content: the same articles, links, figures, bold and italics as the original. No information is added or removed, except the explanation of anything specific to Spain (below).
- Names are not translated: products, apps, companies and the three groups (“Vibe Coding Educativo”, “EduMedia-IAG”). The characters in the cartoon do have a name in each language (
scripts/idiomas.py). - Official material from elsewhere stays in its original language: the title of a law, a report, a guide. A Spanish body is translated if it has a customary form in the language (the ministry, a regional education department); those with a proper name in their own language are not (Generalitat, Xunta, Ertzaintza). Official acronyms are kept (LOMLOE, INTEF, AEPD).
- What is specific to Spain is explained when readers of that language may not know it, above all in English: an educational stage, the oposiciones, an autonomous community.
- A chart (
grafico) is translated in its title, its labels, the names of its series and its note; the type, the unit and the values are left untouched, and the title of the source stays in its original language. - A source in another language carries its mark (“(en castellà)”, “(in Spanish)”), as
scripts/idiomas.pysets out. - When in doubt, consult the language's reference work (in its section) rather than settling it from memory. If a work cannot be read, look for the same answer in another one on the list.
- The same term every time: the AI vocabulary of each language is fixed below. Forms are not alternated (“xatbot” and “chatbot” in the same issue).
- What is correct is not corrected: if a form looks odd but the standard records it, it is left alone.
Catalan
Standard: the standard of the Institut d'Estudis Catalans (IEC), in journalistic register.
Where to look
- Optimot: searches at once the IEC dictionary, the query sheets and the TERMCAT terminology. It is the route to TERMCAT, whose website cannot be read by programs.
- Diccionari de la llengua catalana (DIEC2), Gramàtica (GIEC) and Ortografia of the IEC.
- ésAdir and the Llibre d'estil de 3Cat: the journalistic style of public radio and television.
Format
- Times with a full stop: “a les 10.30”.
- Percentage with a space: “10 %”, as Optimot requires.
- Thousands with a full stop (115.000); ordinals “4t d'ESO”, “1r de batxillerat”.
- Job titles, educational stages and subjects in lower case (“primària”, “batxillerat”); bodies in capitals.
- Place names in their traditional Catalan form: Saragossa, Osca, la Corunya, Biscaia.
Vocabulary
- “el professorat”, “l'alumnat”; “soc”, without an accent.
- AI: “intel·ligència artificial (IA)”, “IA generativa”, “model de llenguatge gran (LLM)”, “agent d'IA”, “aprenentatge automàtic”, “al·lucinació”, “codi obert”, “aplicació”.
- Prompt: “indicació”, the term standardised by TERMCAT; the first time, with the English in italics in brackets: “una indicació (prompt)”.
- Chatbot: “xatbot”.
Calques to avoid: “en base a”, “a nivell de”. Correct, left unchanged: “degut a”, “sí que”, “app”.
Galician
Standard: the rules of the Real Academia Galega (RAG), in journalistic register.
Where to look
- Dicionario da RAG and the Normas ortográficas e morfolóxicas.
- Na Universidade en galego sen dúbida and the Xunta's course in administrative language: numbers, capitals, place names and the position of the pronoun.
- Digalego (dictionary with equivalents), TERGAL (terminology) and the Nomenclátor (place names in Galicia).
Format
- Times with a full stop: “ás 10.30”.
- Percentage with a space: “30 %”.
- Thousands with a space: “115 000”.
- Job titles, days, months and stages in lower case: “educación primaria”, “bacharelato”, “educación secundaria”.
- Place names in Catalonia and the Basque Country in their official form: “Girona”, “Donostia”.
Vocabulary
- “o profesorado”, “o alumnado”; mind the gender (“o xiz”) and avoid Spanishisms (“encerado”, not “pizarra”; “relatorios”, not “ponencias”).
- AI: “intelixencia artificial (IA)”, “IA xerativa”, “modelo de linguaxe (LLM)”, “axente de IA”, “aprendizaxe automática”, “código aberto”, “aplicación”. The last four are not in the RAG dictionary with this meaning: the paper uses them that way.
- Prompt, in italics, and explained the first time as “instrución”, as the Xunta does.
Grammar that often goes wrong
- The pronoun goes before the verb when it is preceded by an adverb or a negation with no pause: “só se equivoca”, not “só equivócase”.
- “Ambos” takes an article before the noun: “ambos os modelos”.
Basque
Standard: standard Basque (euskara batua) of Euskaltzaindia, in journalistic register.
Where to look
- Rules of Euskaltzaindia (each one as a PDF) and its dictionary.
- Euskalterm (terminology), Elhuyar and Labayru (dictionaries).
- Berria's style book: journalistic style.
- Kalkoen Behatokia of the UPV/EHU: calques from Spanish.
Format (Euskaltzaindia rule 203)
- Times: “10:30ean”. Dates: “2026ko azaroaren 13an”.
- Percentage with the sign first and a space: “% 35”.
- Thousands with a space: “115 000”.
- Job titles in lower case; bodies and educational stages in capitals (“Lehen Hezkuntza”, “Batxilergoa”), as rule 200 requires.
- Acronyms and names declined without a hyphen: “AAren”, “ElevenLabsen”.
- Place names: rules 32 (Spain), 151 (regions) and 38 (countries).
Vocabulary
- “irakasleak”, “ikasleak”; the paper is “egunkaria”, never “diarioa”; a school is “ikastetxea”, not “zentroa”.
- AI: “adimen artifiziala (AA)”, “AA sortzailea”, “ikaste automatikoa”, “kode irekia”, “aplikazioa”, “txatbota”. The language model and the AI agent have no standardised term: the paper uses “hizkuntza-eredua” and “AA agentea”.
- Prompt, in italics, untranslated: it has no standardised term.
Calques to avoid: the unnecessary “bere” (Kalkoen Behatokia).
English
Standard: British English, in journalistic register. The paper is “the paper”, never “the diary”.
Where to look
- The Guardian style guide (each letter on its own page) and the BBC one.
- The European Commission's English Style Guide for translators, with the Country Compendium: Spanish bodies and place names.
- IATE (EU terminology) and the Oxford Learner's Dictionary.
Format
- Dates: “13 November 2026”. Times: “10.30am”, as The Guardian has it.
- Percentage attached to the figure: “35%”. Thousands with a comma: “115,000”.
- English quotation marks (“ ”), never the Spanish ones; -ise spelling (“organise”).
- Job titles, stages and generic bodies in lower case (“the education ministry”, “secondary education”).
- Place names: A Coruña, Lleida, Catalonia, the Basque Country.
Vocabulary
- “teachers”, “students” or “pupils”.
- AI: “AI”, “generative AI”, “large language model (LLM)”, “AI agent”, “machine learning”, “hallucination”, “open-source” (hyphenated before a noun), “app”.
- What was said in English in the original is copied from the source, not retranslated.
What is specific to Spain
- A body, with its descriptive name in English and the original acronym: “the Spanish education ministry”, “Spain's education law (LOMLOE)”; never an invented English acronym.
- “autonomous community” (comunidad autónoma), “regional education ministry” (consejería), “competitive civil service exam” (oposiciones), “vocational training” (FP).
- Never “Year 10” or any other equivalent from the British system: the school years do not match. The Spanish name with the age: “the fourth year of ESO (students aged 15 to 16)”, “Bachillerato (the final two years of secondary school, ages 16 to 18)”.