eduKate Singapore

VIEW THIS AS

Auto mode follows the Route Engine until you choose a viewpoint.

YOU ARE HERE

ROUTE CHECK

CONNECTED TO

WHAT NEXT

Use the canonical route for this room, or HELP if you are unsure.

Next useful routeCanonical hubHELP

How People Translate Quickly | LaTeX Mathematics Workbook Translation QA: Equations, References, Units and Multilingual PDFs

—

by

eduKate Asia

Translating a LaTeX mathematics workbook is a precision exercise disguised as a writing task. A question may contain ordinary English, a mathematical expression, a diagram reference and a worked solution. Change the words carelessly and a correct algebra problem can become a different problem; change the TeX control sequences carelessly and the document may stop compiling altogether. If you need to translate a LaTeX document, localize a mathematics worksheet or prepare a multilingual LaTeX PDF, the fastest reliable method is to separate what the student should read from what the typesetting engine and mathematics must preserve.

Start with a frozen source, identify translatable prose and protected notation, translate one representative section, then compile and compare both a mathematical checklist and the rendered pages. Preserve values, variables, operators, equation numbering, labels, references, units and answer relationships unless the curriculum owner has authorized an adaptation. Test multilingual fonts and language settings early rather than discovering an unsupported glyph after translating an entire workbook. A successful build is one checkpoint, but not the finish line.

This guide follows a single reader job: turn an existing, correct LaTeX maths workbook into a target-language workbook without changing what its exercises assess. We will use invented problems rather than copying school questions; show the difference between prose, macros and mathematics; examine the babel and polyglossia language routes; and work through common failures, practical validation, a transfer exercise and publication checks. The output may serve students in different school systems, but the translation must not silently claim curriculum alignment that nobody verified.

Contents: choose the section you need

The basic distinction: language is wrapped around mathematics

Consider the question: “A rectangle has length 12 cm and width 7 cm. Find its area.” In a LaTeX file, part of that sentence may be plain text and the dimensions may be written as $12\,\mathrm{cm}$ and $7\,\mathrm{cm}$. The target language can reorder the sentence substantially while the mathematical invariant remains 12 × 7 = 84, with area measured in square centimetres. The translator should not alter the values or replace the area task with perimeter simply because the target phrase sounds more natural in a familiar classroom idiom.

The LaTeX source has at least three layers. The first is user-visible prose: prompts, headings, instructions, hints, explanations and answers. The second is mathematical content: symbols, equations, numeric examples, units, constraints and diagrams. The third is document machinery: commands, environments, labels, cross-references, package options and file inclusions. Most translation happens in the first layer; meaningful verification checks the second; preservation checks protect the third. An article about ordinary document translation that ignores this separation is not enough for a working mathematics source file.

It is tempting to treat every letter sequence as a word. But \label{eq:quadratic} is a reference target, not a caption to translate casually. \ref{eq:quadratic} is a reference lookup that must still resolve. The \begin{align} and \end{align} markers define the structure of displayed mathematics. A macro name such as \frac has an implementation meaning; its appearance on the screen is not an invitation to translate its spelling. Replacing the surrounding sentence while preserving the command is appropriate. Replacing the command name with a translated word generally is not.

There are harder boundaries. A variable named x is a mathematical symbol, but the source may introduce it with the sentence “Let x denote the number of tickets.” Translate the definition carefully while retaining the variable used by the equation. If the exercise relies on a label like A for a point on a diagram, keep that relationship unless the whole figure and solution are approved for coordinated change. The student’s learning depends on consistent naming across problem, figure and answer.

The quickest quality decision is to make a protection inventory before a translation begins. List which objects cannot change: coefficients, signs, units, labels, cross-reference keys, diagrams, equation identities, and final answer values. Record any permitted local presentation choices separately: decimal notation, punctuation, words for mathematical operations, translated heading labels, or font and page-layout adjustments. That makes it easier to distinguish faithful localization from an unapproved revision to the mathematics.

Freeze a trustworthy source and know what belongs to it

A TeX workbook may span several files: a main .tex, chapter includes, figures, bibliographies, custom style files and a set of answer files. Translating the main source alone may leave exercises in an included chapter untouched. Before editing, compile the source as received and keep a copy of its generated PDF. Record which TeX engine and distribution were used, which file is the root document, and whether there were warnings that already existed. This baseline stops an old formatting defect from being incorrectly attributed to the translation.

Review the project structure with the curriculum or publishing owner. Which files contain actual student tasks? Which contain reusable macros? Is there an answer key generated from a shared data source or separately typed prose? Is an optional chapter excluded by a build flag? A translation map should include the paths of files in scope, not just the visible title of the workbook. Someone who checks only the first ten pages may miss an appendix, a scoring rubric or a hidden teacher edition.

Create the target language in a branch or separate copy, and keep the source language version immutable. If a workbook is edited frequently, establish a simple version key, such as a source commit or dated approved edition, so translators can determine whether a question changed upstream. A translation based on last week’s numbers may be wrong today even if it was perfect when written. Synchronization is especially important where a teacher adjusts coefficients or corrects answers before examinations.

Build a content inventory by exercise ID. A representative row could contain “Chapter 2, Question 8, perimeter of rectangle, two numeric givens, one diagram, one answer paragraph, one reference to Figure 3.” This lets reviewers compare the same question across languages without depending on page numbers, which often shift after translation. The inventory also exposes missing segments: if the source has eighty numbered exercises and the target has seventy-nine, you have a completeness problem regardless of grammatical quality.

A useful baseline captures the number of questions, subquestions, figures, displayed equations and solutions. These numbers are not proof of equivalence—one can preserve the count while corrupting a minus sign—but discrepancies are excellent diagnostic triggers. Define what may legitimately differ, such as the number of pages or lines, and what should not, such as exercise identities and answer mapping. Keep screenshots or reference PDFs for complex figures and tables.

Separate TeX control sequences from translatable prose

Not every word in a .tex file is printed. Comments beginning with % may contain guidance to authors, while macro definitions control style or content in multiple places. Some commands, such as \section{Fractions}, have an argument that is visible text and should be translated; the command name itself should remain intact. Other arguments contain labels, paths or options that should not change. The editor must understand the context of an argument before replacing its contents.

For example, in \section{Adding Fractions}\label{sec:fractions}, “Adding Fractions” is the student-facing section title while sec:fractions is a stable reference key. Translating the visible title and preserving the key is normally the safe choice. In another location, \ref{sec:fractions} will continue to produce the correct section number after compilation. If the key is changed at one site but not at all references, the document may produce an unresolved reference. A bad cross-reference can misdirect students to the wrong explanation even though the mathematics itself is fine.

The same principle applies to tables. Column headings are prose; alignment operators like &, row terminators \\, and environment delimiters are structure. A translated explanation may contain characters that have special meanings to TeX, such as %, &, _, # or a backslash. Entering them naively can change the markup or provoke compilation errors. Use appropriate escaping or text commands, and verify the actual rendered result. This is one reason direct replacement in source files demands stronger QA than copy-editing a word processor document.

Mathematics mode has its own conventions. $x+2=5$ is a mathematical expression, whereas “subtract two from both sides” is prose. Copy the expression unchanged unless there is an authorized mathematical correction. If the target-language sentence has a different order, move the math token as a whole rather than editing characters inside it blindly. For larger expressions, a segment map can identify which surrounding words connect to the formula, preventing an explanation from reversing “greater than” or “less than.”

Avoid treating every visible word as translatable in isolation. The command \text{if } inside an aligned derivation might contain English that should be localized, but its context belongs to the proof, not to a menu heading. A callout generated by a macro may need localization at one definition point rather than hundreds of repeated instances. Find the owner of the visible output before making a change; it saves both time and repetition.

Make the worked example your acceptance test

Suppose an original problem says, “A bus travels 180 km in 3 hours at a constant speed. Find the speed in kilometres per hour.” The calculation is 180 ÷ 3 = 60 km/h. The translator can reorganize the sentence for the target language, but three elements must remain: distance 180 km, time 3 h, and constant speed. Removing the condition “constant” might broaden the claim; changing the requested unit to metres per second without converting the answer would make the worksheet inconsistent.

Write the equivalence check as a brief proof. Given quantity: distance is 180 kilometres. Given duration: three hours. Relationship: speed equals distance divided by time under the stated interpretation. Expected answer: 60 kilometres per hour. Compare the source and target prompts against this proof. If the target says the bus covers 180 km each hour, the final answer changes dramatically, and the mistranslation may not be obvious to a language editor unfamiliar with the arithmetic.

Take a second example involving a negative sign: “Solve -2x = 14.” The correct answer is x = -7. If a hyphen or math minus is lost during source manipulation, the target may display 2x = 14 while keeping the original answer. An equation count check will not catch the semantic corruption. A disciplined reviewer solves the problem from the rendered target page without looking at the source solution first, then compares results. This is an effective independent mathematical QA method because it resists confirmation bias.

A third example tests units. The original says a garden is 4 metres by 3 metres and asks for the area. The answer is 12 m², not 12 m and not 14 m². If a translated instruction uses a word normally associated with perimeter, students may correctly compute 2(4+3)=14 m and be marked wrong by an area answer key. The mathematical vocabulary therefore needs contextual review. Translating “area” and “perimeter” by sound or casual dictionary proximity is not enough.

Now consider fractions: “Find three quarters of 28.” The expected computation is 28 × 3/4 = 21. In a TeX expression, the denominator is often wrapped inside \frac{3}{4}. If the translator accidentally changes a brace or number while editing adjacent prose, the PDF may show a different expression or the build may fail. Protect the fraction macro, then verify its rendered numerator and denominator. This is faster than investigating a wrong answer after dozens of pages have been delivered.

These cases illustrate why a workbook needs two reviewers or at least two review passes. One pass checks meaning and register; the other checks the mathematics independently. Both should examine the actual target output, not merely the source .tex. The final PDF is what the child will read, and the final answer key is what will determine whether their work is judged correctly.

Configure language, fonts and engines early

LaTeX can typeset many languages, but the correct route depends on the script and the project environment. The babel package supports multilingual typesetting, language-sensitive hyphenation and other locale conventions across different TeX engines. polyglossia provides another route, commonly used with XeLaTeX or LuaLaTeX. Current CTAN package pages and Overleaf guidance are useful references, but they do not prove that any arbitrary font or existing workbook class is compatible with every target script.

Test a small representative page before translating hundreds of exercises. Include a heading, a normal paragraph, all target-language diacritics or script characters, one mathematical display, a table, and a cross-reference. If the language is right-to-left, include mixed text and formulas. Compile with the intended engine and check whether glyphs appear, line breaking works, headings are placed sensibly, and the mathematics remains readable. A project that fails this smoke test needs a typesetting solution before bulk translation.

Font choice matters even when the document compiles without errors. A missing glyph may be replaced by a blank box or another fallback symbol. A font may support the language’s alphabet but lack a particular mathematical sign or symbol used in the workbook. Switching text fonts should not arbitrarily change the appearance of mathematical variables or confuse Greek letters with Latin lookalikes. Inspect characters such as minus, multiplication, inequality, degree signs, subscripts and superscripts in final PDF pages.

Do not overgeneralize from one script. A Spanish target edition may be handled adequately with familiar font and language settings; an Arabic or Tamil edition can need different shaping, fonts, text direction and engine choices. An East Asian edition may require appropriate CJK typesetting packages or settings. The source project may already contain custom macros that assume left-to-right layout. Verify those assumptions with real pages rather than promising a universal command that solves every language.

The test page should be preserved as part of the build record. If a later package or font update changes output, regenerate the test page and compare it to the approved version. That is an inexpensive way to detect an environment change before it damages a complete localized workbook. It also separates a target-language font defect from a mistranslated exercise, reducing the number of people needed to diagnose the fault.

Handle decimal conventions, number formatting and units carefully

Different languages and regions use different conventions for decimal points, decimal commas, grouping spaces and punctuation. That does not mean a mathematics translator can freely change notation inside the exercise. The learning objective may concern the notation itself. If a worksheet teaches interpretation of 1.5 and 1,500, the source format may be an object of instruction, not merely a locale preference. Identify whether each number is a mathematical value, a formatting example, a thousands separator, a decimal separator or a literal code-like token.

For a worked example, suppose the source question uses 1.25 m for length. If the target edition conventionally writes the decimal fraction with a comma, the corresponding displayed value may be 1,25 m when permitted. But the value remains one and a quarter metres, not one thousand two hundred and fifty metres. The reviewer should calculate from meaning, not from punctuation alone. If a target school’s syllabus prescribes a notation convention, link to that actual syllabus rather than asserting it from general language practice.

Units must remain dimensionally appropriate. Translating “kilometres per hour” does not transform speed into distance, and “square centimetres” cannot be shortened to centimetres without changing the quantity. Mathematical notation such as m^2 or \mathrm{cm}^2 can be preserved with localized surrounding words. When converting units is expressly authorized, show the conversion and update every associated problem and solution consistently. Otherwise keep the given units and correct numeric answers.

Tables and word problems deserve a special check. A translated table header such as “time (min)” may wrap to two lines, causing the numerical columns to appear offset visually. A longer target phrase can also shift a unit into the next line and separate it from its quantity. The PDF may technically contain the right numbers but invite the wrong association. Compare table cell headers, row labels, units and totals after typesetting; do not assume that syntactic validity guarantees visual clarity.

Currency, date and measurement contexts can also create hidden curricular changes. An English problem involving dollars might translate directly in a multilingual edition without replacing the currency. Localizing the scenario to a different country, changing the currency or changing a conversion factor is a separate curriculum adaptation. Make that decision with the owner and then recompute all values. A translation-only task should not pretend that a fictional local scenario remains the same mathematical exercise if the amounts have been quietly altered.

Preserve labels, references, figures and answer mappings

Long workbooks are full of cross-references: “see Figure 4,” “use Equation 2,” “compare with Exercise 12,” and “answers are on page 68.” In TeX, many of these can be generated by stable labels rather than typed numbers. A translation should preserve those labels and verify they resolve after the target edition compiles. The text around the reference may be translated; the key within \ref{...} normally stays the same.

Page numbers are particularly dangerous. The translated edition may have different line lengths and therefore different page breaks. A manually typed instruction “see page 68” can become wrong even when the document compiles. Replace hard-coded page references with suitable supported reference mechanisms where the editorial contract allows, or at minimum recheck them against the final PDF. Do not silently change references by guessing where the new section might land.

The same reasoning applies to figure labels. A geometric drawing could mark points A, B and C while the prose says “Triangle ABC is isosceles.” If the translator changes one point label in the sentence but not the figure, the child encounters a contradiction. Inspect the rendered figure and the corresponding instruction side by side. If diagrams use packages such as TikZ, protect the geometry code and change only explicitly owned display labels. A path coordinate is not a translatable string.

Answer keys must remain aligned with question IDs. A workbook that uses shared macros or external files may generate solutions separately from exercises. Translate the explanatory solution prose but verify every mathematical result and item number. It is helpful to maintain an answer ledger listing problem ID, source expected result, target expected result and independent recalculation. If a solution uses a translated variable definition, check that all later steps still use the same symbol.

Cross-references sometimes require more than one compilation pass to settle. Treat unresolved references and missing labels as build warnings requiring investigation, not cosmetic noise. Compare logs against the baseline, and inspect generated references in the PDF. A document that contains placeholder question marks instead of a figure number is not release-ready just because its pages exist.

Translate reasoning, not merely commands

Mathematical explanations often contain conditional logic that determines why a solution is valid. Words such as “if,” “only if,” “unless,” “at least,” “at most,” “positive,” “nonzero,” “distinct” and “approximately” carry technical force. A mistranslated qualifier can reverse an entire proof while preserving all printed equations. Record these terms in a controlled glossary for the particular target audience, and validate them with examples rather than assuming one dictionary definition fits every classroom context.

Suppose the source says: “A number is divisible by 6 if it is divisible by both 2 and 3.” If “both” becomes “either,” the translated statement is false. Consider the number 4: it is divisible by 2 but not 3, so it is not divisible by 6. The source and target can contain identical numerals yet express different reasoning. A quick counterexample is an excellent QA technique for conditional statements. Test edge cases where an imprecise connective would visibly fail.

Another example: “Solve for nonnegative integer values of n.” This is not the same as “solve for positive integers.” The first includes zero; the second excludes it. An answer key that contains n = 0 may look strange in the target edition if the prompt was mistranslated. Check domain words against the actual solution set. The same applies to “greater than” versus “greater than or equal to”; the difference at the boundary is pedagogically central.

Definitions should be introduced before their advanced uses. If the source calls a rate of change “gradient,” the target edition may use a locally standard equivalent, but it should not arbitrarily alternate among unrelated terms for slope, derivative and incline. Record approved terminology at the chapter level and review reuse in questions, worked solutions and revision summaries. Consistency helps the learner build a stable concept; it should never become a pretext for making every occurrence of an ambiguous word identical.

A longform educational translation succeeds when the student can follow the argument, not merely recognize the symbols. Read a worked solution in the target language and ask whether each “therefore” is earned by the previous step. If a sentence says “divide both sides by x,” examine whether the original assumed x ≠ 0. If that condition disappeared during translation, the explanation can become mathematically unsafe even though the printed algebra still looks formal.

Build source-level and PDF-level regression tests

A practical regression system contains multiple layers. First, parse or compile the LaTeX source with the intended engine. Second, compare protected commands and labels against the source, allowing only documented changes. Third, compare exercise IDs and answer mapping. Fourth, visually inspect a sample covering every problem type and all complex layouts. Fifth, solve a carefully selected sample of target questions independently. Sixth, run the final PDF through whatever accessibility and document checks are appropriate to the actual distribution environment.

A simple text diff can reveal accidental changes to numeric literals or \label keys, but it will also show expected translated prose. Configure or manually interpret the diff to focus on protected regions rather than treating every change as an error. Conversely, identical equation text can still be misinterpreted by a translated instruction. No purely mechanical comparison can certify the whole educational job. Use structured checks to narrow the review and human reasoning to evaluate meaning.

Coverage should be risk weighted. High-risk examples include sign changes, multi-step proofs, word problems involving units, diagrams with text labels, questions with conditions, and answer keys for multiple choice. A chapter of independent numeric drills may need a different sampling depth from a chapter teaching logical implication. Report the number and types of items actually checked; do not invent a 100% proof of correctness from a script that only counts math environments.

For a long workbook, a table can connect a defect to its location: question ID, page in target PDF, source file, error category, severity, fix owner, and retest result. A wrong answer key is a release blocker. A missing diacritic in a heading may also be a blocker if it changes an important word. A shifted page break is often acceptable; an equation clipped at the right edge is not. Severity follows educational consequence, not simply how easy the bug is to repair.

An especially revealing final check is to provide target-language pages to a mathematically competent reviewer who has not read the source and ask them to solve selected problems. Compare the independently obtained answers with the key, then trace any discrepancies. This method catches contradictions that a bilingual reviewer may unconsciously correct in their head. It also tests whether the instructions are genuinely understandable without access to English.

Common failures and what they tell you

A build error immediately after translation often indicates damaged TeX syntax, a special character not escaped correctly, an invalid command or unsupported character in the selected engine. Start with the compilation log, isolate the first error, and check the changed segment. Do not retranslate whole paragraphs when a single unescaped symbol is the cause. A missing glyph in the PDF indicates a font or engine problem, not automatically a wording problem.

An unresolved ?? near a figure reference suggests that labels or build passes require attention. Compare the source keys and target references. A table overflows because translated headings grew; adjust layout without changing numerical values. A multiple-choice answer no longer matches its label because option letters were reordered; restore the intended mapping. A solution is mathematically inconsistent with the question because an inequality sign or unit changed; rederive the answer and repair the semantic mismatch before release.

A more subtle defect appears when a source definition is translated one way in the introduction and differently in worked examples. Students may think two terms describe different concepts. Check the glossary, especially terms that carry a boundary such as “ratio,” “rate,” “proportion” and “percentage.” Precision matters more than rhetorical elegance when a child is learning a new operation for the first time.

Some changes are legitimate adaptations rather than errors. A local edition may change a scenario from miles to kilometres, adapt an example currency, or rename a school-year category. But such changes require mathematics, curriculum and editorial approval. They also require rechecking every calculation. Track adaptations separately from translation so a later reviewer does not assume the two versions differ only in language.

Do not conceal pre-existing source errors. If a workbook contains an incorrect answer before translation, preserve the report, notify the content owner and get a separate decision about correction in both languages. Silently “fixing” only one edition may create a discrepancy in instruction or assessment. The best action is to make the source truth and target truth explicit and traceable.

A release process that protects the learner

Create a final acceptance route: compile from a clean environment; confirm fonts and language settings; reconcile question and solution counts; recheck labels and cross-references; render each special problem type; verify units, signs, domain conditions and answers; and read representative instructions with a target-language reviewer. Save the finished PDF alongside the exact .tex source that produced it. Keep the unchanged source edition for audit and for future updates.

The release notes should identify language, source edition, compiler, important packages and any deliberate adaptations. If the target edition uses a different assessment convention, state that explicitly. If accessibility checking was limited to manual visual review, say so rather than claiming compliance with standards never tested. Transparent scope helps teachers choose appropriate materials and helps the next editor understand what can be reused safely.

When the source workbook changes, process the change as a new content event. Do not automatically re-import earlier translations into modified algebra without reviewing the new numbers and solution relationships. A translation memory is useful for stable instructions and terms, but mathematical correctness is attached to a specific exercise version. The safer workflow is selective reuse followed by targeted verification of changed problems.

A compact learner check is also useful. Ask whether the child can identify the given quantities, the requested unknown, the constraints and the expected unit from the target-language question alone. If those four components remain clear, many common semantic failures will be caught early. This mirrors good mathematics teaching: understanding comes before computation.

Transfer challenge: two translated questions that look equally plausible

Imagine a source worksheet with two questions. Question A: “Find the area of a square with side length 5 cm.” Question B: “Find the perimeter of that square.” The source answers are 25 cm² and 20 cm. A target-language reviewer receives a translation in which both questions use the same ambiguous word for “size,” though all numbers and formula markup are unchanged. Does the translation pass? No. The correct method cannot be recovered unambiguously from the target instructions, and the units in the answer key reveal that the intended quantities differ.

Repair the problem by selecting precise target-language terms for area and perimeter, then have a reviewer solve each question from the translated prose. Confirm the answer units: 5 × 5 = 25 square centimetres for area; 4 × 5 = 20 centimetres for perimeter. Check that the localized font renders the squared unit correctly. A table of source symbols alone would miss the original failure because the formulas were never damaged.

Now change the condition: the workbook asks for the length of a square’s side given its area of 25 cm². The calculation is the square root, giving 5 cm. A translator who reuses the former area question’s wording can accidentally make students compute the area again. The numeric values look familiar, but the direction of inference is reversed. The transfer answer is to identify the unknown and the operation required by the new prompt before reusing a phrase from the older translation.

The same practice extends to reading comprehension and Science: record what is given, what is asked, what relationship connects them, and what would count as an answer. It is an educational method as much as a localization method. Quick translation becomes dependable when students’ reasoning is used as the ultimate check on the words.

Summary: translate prose, defend the mathematical contract

LaTeX is powerful because a workbook’s prose, mathematics and document structure can be represented precisely. That precision also makes careless string replacement dangerous. Translate the parts students must read, preserve control sequences and mathematical relationships, configure language support early, compile often, then verify the final PDF and answer key against independent reasoning. When the output has the right language but the wrong question, the translation has failed.

A strong review can be performed in stages rather than as one overwhelming proof: source inventory, protected-token check, language and script test, worked examples, cross-reference check, independent answer solving and release packaging. Those stages make the work faster by identifying where an error belongs and what needs retesting after a repair.

Frequently asked questions

Can I translate everything between braces in a .tex file?

No. Braced content may be visible prose, a reference key, a filename, a numerical argument, an option or a macro definition. Understand the command and its context before changing its argument.

Should I translate \label{} keys and \ref{} targets?

Normally preserve stable reference keys so the target edition’s cross-references continue resolving. Translate the visible text around the reference instead. Any deliberate key migration requires synchronized updates and a fresh compile check.

Is a successful PDF build enough to approve the maths?

No. A PDF can compile perfectly while a minus sign, quantifier or unit has been mistranslated. Independently solve representative target-language problems and compare answers with the key.

Do I need babel or polyglossia?

You need an appropriate language and font setup for the intended script and compiler. babel and polyglossia are established options with different engine and project considerations. Consult current documentation and test a representative page before full production.

What if the target font cannot display a character?

Check the actual TeX engine and font coverage, then use a supported font and appropriate packages or compiler. Do not substitute visually similar mathematical symbols without a meaning check.

May I change a decimal point to a comma?

Only when that change is appropriate to the target context and does not change the learning objective or numeric interpretation. Verify the displayed value and calculations; some exercises specifically teach number notation.

What if an answer key is wrong in the source?

Report it separately to the content owner. Coordinate a correction for affected editions and recompute the problem; do not silently create a target-language-only change under the name of translation.

How do I handle diagrams created in TikZ?

Protect coordinates, shapes and mathematical labels, and translate only those visible words within the approved scope. Render the actual target diagram and check that it still matches instructions and solution references.

Why do page numbers change after translation?

Target-language text may wrap differently and change pagination. Recheck hard-coded page references and prefer supported dynamic references when the editorial system allows them.

What should accompany the final files?

Include the exact target .tex sources, generated PDF, language and compiler details, a change log, the reference source edition, and a validation record describing what was actually checked.

Authoritative references and further reading

Suggested internal reading routes (verify destinations before WordPress publication)

  • Link to eduKateSG’s existing Mathematics Learning System resources at independently verified canonical URLs for prerequisite reasoning and unit checks.
  • Connect to the separately prepared MathML accessibility article for spoken equation delivery, but only when its publication URL can be verified. It has a different reader job from translating TeX source.
  • Where appropriate, point readers to verified vocabulary-learning guides for precise mathematical terminology rather than duplicating the underlying vocabulary pedagogy.