Proofreading Cyrillic and Latin Texts

Proofreading Serbian requires two different kinds of attention. A script pass catches mixed alphabets, lookalike characters, lost diacritics, and faulty conversion; a language pass catches case endings, clitic placement and spacing, capitalization, punctuation, and inconsistent standard variants. A text can look visually Serbian while failing either pass.

The safest workflow is systematic: decide the intended script and variety first, then check one error class at a time. Reading the whole text repeatedly for “anything wrong” is slower and misses patterned errors.

Pass 1: Confirm the Intended Script

Serbian Cyrillic and Latin are both standard scripts, but an ordinary word should not switch alphabet halfway through. Copying from websites, spreadsheets, and OCR often inserts characters from the other script that look identical on screen. Cyrillic а, е, о, с, х closely resemble Latin a, e, o, c, x, while Latin j resembles Cyrillic ј but has a different code point.

These are equivalent Serbian sentences. Choose one script for the running text unless a quotation, registered name, URL, code sample, or other real-world item requires another.

Mixed-script errors can be invisible to the eye but visible to search, sorting, text-to-speech, and spell-checking. Do not rely only on visual similarity. In a Cyrillic document, select a suspicious word and inspect its language or character information; in a Latin document, search for Cyrillic code points.

Pass 2: Check the One-to-One Conversion Units

Most letters convert directly. The main traps are Cyrillic љ, њ, џ, which correspond to the Latin digraph letters lj, nj, dž, and Cyrillic ђ, which corresponds to đ, never plain dj.

The Latin conversion is Ljiljana je Đorđu poslala kartu Ljubljane. Sentence-initial digraph letters use Lj/Nj/Dž, not LJ/NJ/DŽ, unless the whole word is in capitals.

Automatic converters need lexical awareness because the character sequences nj, lj, and dž do not always represent digraph letters: they can straddle a morpheme boundary or represent separate sounds in learned and foreign words. Injekcija must become инјекција, not ињекција. Proofread the output rather than trusting a global replacement.

Pass 3: Restore Every Latin Diacritic

Standard Serbian Latin distinguishes c–č–ć, z–ž, s–š, and d–đ. Removing the marks is common in hurried messages, but it creates nonstandard spelling and can erase lexical distinctions. Search separately for suspicious plain sequences; a spell-checker cannot always infer the intended word.

Do not “repair” đ as dj or dž. The three spellings represent different structures: ђ/đ, д + ј/d + j as in одјек/odjek, and џ/dž as in џеп/džep.

Pass 4: Verify Word Boundaries and Clitics

Spacing survives script conversion unchanged, but it still needs grammatical proofreading. Regular verbal negation is separate (не знам / ne znam), the question particle is separate (да ли / da li), and clitics are written as separate orthographic words (сам га видео / sam ga video). Certain lexicalized negatives such as нећу / neću, немам / nemam, and нисам / nisam are joined.

The cluster сам га has a natural second-position placement and the order auxiliary + pronoun. Do not fuse it as самга, and do not move a clitic merely to imitate English word order.

Future I needs an additional spelling check: писаћу, радићемо are joined after infinitives in -ти, while доћи ћу, наћи ћемо remain separate after -ћи.

For the full boundary logic, see writing words together or separately.

Pass 5: Check Case Endings and Agreement

A perfectly converted word can still have the wrong grammatical ending. Prepositions govern cases, adjectives agree with their nouns, and masculine animate direct objects have genitive-like accusative forms. Proofread phrases as units rather than checking each dictionary stem separately.

Here са requires instrumental добрим пријатељем, while о requires locative новом послу. The forms са добар пријатељ and о нови посао are not spelling variants; they are case errors.

Masculine animate колега takes accusative колегу, and its adjective takes новог. A proofreading tool that only recognizes individual words may accept a mismatched phrase, so agreement must be checked by a human reader or grammar-aware system.

The numeral, adjective, and noun form one agreement domain. Read from the head noun outward and confirm number, gender, and case.

Pass 6: Keep Ekavian or Ijekavian Consistent

Ekavian and Ijekavian are both standard in Serbian. Proofreading should respect the declared model rather than rank one as better. This guide uses Ekavian as its productive baseline, so an Ekavian document should normally keep sets such as време, недеља, млеко, проверити, and извештај together.

The standard Ijekavian Serbian counterpart is Састанак је сљедеће недјеље, па ћемо извјештај провјерити на вријеме. That sentence is valid when Ijekavian is intended. An editor should not silently mix следеће недјеље or извјештај проверити inside one neutral model unless quoting or deliberately comparing forms.

Lexical choices associated with neighboring BCMS standards require the same factual treatment. If a Serbian Ekavian house style calls for хиљада and извештај, a Croatian quotation containing tisuća or izvješće must remain a quotation; it should not be relabeled as “bad Serbian” or silently rewritten without editorial authority.

Pass 7: Capitalization and Punctuation

English title capitalization should not be copied into Serbian. In ordinary Serbian titles and institutional names, capitalize the first word and any independently proper-name element, not every important word.

Serbian quotation marks are conventionally „…“, though digital texts may use other consistent styles. The closing punctuation belongs to the sentence structure, not to an automatic copy of English curly quotes. Also check commas before adversative conjunctions such as али and него, and do not insert an English Oxford comma mechanically into a simple Serbian list.

A Full Proofreading Task

The following draft deliberately mixes error types. First label each error as script, diacritic, conversion, spacing, case, variant, capitalization, or punctuation. Then repair it without changing the meaning.

A Latin-only correction is: Ne znam da li je Ljiljana poslala izveštaj novom direktoru, ali proveriću posle sastanka. Compare the two correct versions line by line:

• Љиљана ↔ Ljiljana, not Лјилјана;
• извештај ↔ izveštaj, with ш/š;
• директору ↔ direktoru, preserving the dative ending;
• проверићу ↔ proveriću, joined in both scripts.

For dictation, use the corrected sentence without displaying it. The proofreader should first choose Cyrillic or Latin, then perform all seven passes. A second person can compare the result with the other-script version; differences in spaces or endings reveal that more than transliteration changed.

Common Mistakes

Trusting visual lookalikes

Converting digraphs mechanically

Losing Latin diacritics

Checking spelling but not cases

Mixing variants inside an Ekavian text

Key Takeaways

• Proofread script integrity before grammar; mixed lookalike characters can be visually invisible.
• Convert љ/lj, њ/nj, џ/dž, and ђ/đ as exact units, with context for real n + j sequences.
• Restore every Latin diacritic and keep č–ć and dž–đ distinct.
• Check word boundaries, clitic order, case endings, and agreement after conversion.
• Treat Ekavian and Ijekavian as valid Serbian standards, but keep the chosen editorial model consistent.