The 7-step decision tree for translating character names, sects, locations, and techniques without turning Zhang Wei into Zhang Great or leaving Baihua Building in raw pinyin
Tri Thu知秋
· Updated
On this page
- Why does sorting proper nouns by rigid labels like Person, Place, or Organization break web novel translations?
- How do the 7 steps adapt across English, Vietnamese, Indonesian, Spanish, and Thai reader conventions?
- How do I lock names in the "Names & terms" tab (Step 1) and what are the 4 system limits I should know?
In novel translation, whether to keep a proper noun in phonetic Pinyin or translate its meaning depends on its naming function and descriptive meaning, never on a rigid entity label like Person, Place, or Organization alone. NovelVerba evaluates proper nouns through a 7-step priority tree, starting from locked entries in your Names & terms tab, restoring CJK-transcribed foreign names, and keeping native personal names in Pinyin or Romaji while translating meaningful locations, sects, and techniques.

| Priority step in the decision tree | Rule by naming function and meaning | Concrete English translation examples |
|---|---|---|
| Step 1: Glossary entries win unconditionally | Any term listed in your active glossary always follows the locked mapping: the AI will not independently translate, expand, modernize, or reinterpret it even when the literal meaning looks obvious | Resolved across 3 tiers: Names & terms (per-novel tab) > My dictionary > Genre packs |
| Step 2: Foreign names transcribed into CJK characters | Restored to the original non-CJK name whenever confidently identifiable rather than romanized as Chinese Pinyin | 赫敏 -> Hermione (not Hemin); 哈利·波特 -> Harry Potter |
| Step 3: Native Chinese or Japanese personal names and numbered chapter headings | Follow standard target-language transliteration for personal names and never paste raw CJK script; numbered chapter or story headings are translated fully as regular content | Chinese personal names use tone-mark-free Hanyu Pinyin with surname first: 张伟 -> Zhang Wei, 李白 -> Li Bai |
| Step 4: Incantations, chants, foreign formulae, and Katakana skill coinages | Keep opaque incantations, magic words, and deliberately foreign formulae as-is; however, descriptive spell/technique names and light-novel Katakana skill or item coinages are sent to Step 5 for semantic translation | Keep Hogwarts, Expelliarmus, prae caeteris as-is; treat 破天剑诀 (Heaven-Breaking Sword Art) and Katakana attack names as meaningful techniques under Step 5 |
| Step 5: Descriptive names (places, organizations, titles, techniques, items) | Translate naturally into English when the meaning of the place, sect, organization, title, or technique is part of how readers understand the world | 百花楼 -> House of a Hundred Flowers (not Baihua Building); 天机阁 -> Pavilion of Heavenly Secrets; 忘忧谷 -> Valley of Forgetting; 云剑宗 -> Cloud Sword Sect; 破天剑诀 -> Heaven-Breaking Sword Art; 华夏集团 -> Huaxia Group |
| Step 6: Do not translate the literal dictionary meaning of personal or family names | Conventional personal and family names stay phonetic or original even when their individual characters or words have dictionary meanings; special nicknames belong in Step 1 | Sirius Black stays Sirius Black; 张伟 stays Zhang Wei (never Zhang Great) |
| Step 7: When uncertain between a phonetic name and a semantic term | Prefer the conservative rendering from Steps 2 to 4 and Step 6 and leave the candidate for glossary review, never inventing a wild translation or dropping raw CJK characters into the output | Uncertain candidates land in the unticked Needs review group during Automatic names & terms scan so you decide before clicking Apply |
Why does sorting proper nouns by rigid labels like Person, Place, or Organization break web novel translations?
If you have ever read an English web novel translation where the protagonist 张伟 is bizarrely called Zhang Great, while evocative locations like 百花楼 and 天机阁 are left half-translated as Baihua Building and Tianji Building, you have seen what happens when a translation prompt judges proper nouns by surface entity tags alone. If you tell an AI to keep all Places and Organizations in Pinyin, it turns 百花楼 into Baihua Building and strips away the imagery of the House of a Hundred Flowers, or fails to render institutional names like Ministry of Magic naturally. Conversely, if you tell it to translate every character that has a dictionary meaning, it turns the given name 伟 in 张伟 into Great and translates family names like Black in Sirius Black.
Deciding which form a proper noun should take is a separate problem from keeping a character's name consistent across chapters: first you have to choose the right form for each category of noun. Instead of relying on entity type labels alone, NovelVerba checks every proper noun against a 7-step priority tree based on naming function and meaning. Compare how a single action line reads when handled by rigid rules versus the 7-step policy:
Rigid rule output: "Outside Baihua Building, Zhang Great and Hemin drew their blades to unleash Po Tian Jian Jue of the Yunjian Organization." -> With the 7-step decision tree: "Outside the House of a Hundred Flowers, Zhang Wei and Hermione drew their blades to unleash the Heaven-Breaking Sword Art of the Cloud Sword Sect."
In the second sentence, every proper noun lands in the form English web-novel readers expect: native Chinese personal names stay in clean Pinyin, CJK-transcribed Western names return to their original English spelling, and descriptive pavilions, sects, and martial arts convey their imagery in English. Whenever I set up a new project, I lock these proper-noun forms first and then pair them with relationship-based dialogue pronouns in Choose address terms by character relationship when translating Chinese novels and sect seniority titles in Translating xianxia forms of address: master, martial uncle and senior so both names and spoken titles read naturally.
How do the 7 steps adapt across English, Vietnamese, Indonesian, Spanish, and Thai reader conventions?
While the 7-step priority structure is identical across all eight target languages supported by NovelVerba, Steps 3 and 5 adapt to the established reading conventions of each language community. In English, Indonesian, and Thai, readers rely on semantic translation in Step 5 to understand the imagery behind cultivation sects, martial techniques, and classical pavilions: 云剑宗 becomes Cloud Sword Sect (EN), Sekte Pedang Awan (ID), and นิกายดาบเมฆ (TH), while 破天剑诀 becomes Heaven-Breaking Sword Art (EN), Teknik Pedang Pembelah Langit (ID), and ศาสตร์ดาบทำลายฟ้า (TH). Modern corporate names combine the phonetic brand name with the translated organizational noun, turning 华夏集团 into Huaxia Group in English, Grup Huaxia in Indonesian, Grupo Huaxia in Spanish, and กลุ่มหัวเซี่ย in Thai.
Two target languages have notable community exceptions in Step 5. In Spanish xianxia translations, technique names are translated by meaning (破天剑诀 -> Técnica de la Espada Rompecielos), but sect names keep their Pinyin proper core preceded by the Spanish noun Secta (云剑宗 -> Secta Yunjian, rather than Secta Espada de Nubes). In Vietnamese, readers already know a Chinese-derived layer of vocabulary, so xianxia sect and technique names use the traditional Vietnamese reading of each character (云剑宗 -> Vân Kiếm Tông, 破天剑诀 -> Phá Thiên Kiếm Quyết), while descriptive places like 百花楼, 天机阁, and 忘忧谷 can be locked in the per-novel dictionary whenever a reader prefers the character reading over semantic translation.
| Source term and category | English (EN) rendering under the 7 steps | How VI, ID, ES, and TH handle the same term |
|---|---|---|
| Native Chinese personal name: 张伟 / 李白 (Steps 3 & 6) | Pinyin without tone marks, surname first: Zhang Wei / Li Bai (never translated literally as Zhang Great) | ID & ES: Zhang Wei / Li Bai (Pinyin); VI: Trương Vĩ / Lý Bạch (Vietnamese reading of the characters); TH: จางเหวย / หลี่ไป่ (Thai phonetic script) |
| CJK-transcribed foreign name: 赫敏 / 哈利·波特 (Step 2) | Restored to original Western name when confidently identifiable: Hermione / Harry Potter | VI, ID, ES restore to Hermione / Harry Potter; TH renders the restored foreign name into Thai script: เฮอร์ไมโอนี่ / แฮร์รี่ พอตเตอร์ |
| Cultivation sect name: 云剑宗 (Step 5) | Translated semantically into English: Cloud Sword Sect | ID: Sekte Pedang Awan; TH: นิกายดาบเมฆ; VI keeps the Vietnamese reading of the characters: Vân Kiếm Tông; ES keeps Pinyin core + Secta: Secta Yunjian |
| Meaningful martial technique: 破天剑诀 (Steps 4 & 5) | Recognized as a descriptive technique rather than an incantation: Heaven-Breaking Sword Art | ID: Teknik Pedang Pembelah Langit; ES: Técnica de la Espada Rompecielos; TH: ศาสตร์ดาบทำลายฟ้า; VI: Phá Thiên Kiếm Quyết |
| Descriptive classical locations: 百花楼 / 天机阁 / 忘忧谷 (Steps 1 & 5) | Translated naturally by meaning: House of a Hundred Flowers / Pavilion of Heavenly Secrets / Valley of Forgetting | ID, ES, TH translate the descriptive meaning; in VI, readers can lock the Vietnamese character readings (Bách Hoa Lâu, Thiên Cơ Các, Vong Ưu Cốc) in Names & terms (Step 1) |
| Urban group & Western proper nouns: 华夏集团 / Sorting Hat / Ministry of Magic / Diagon Alley / Sirius Black (Steps 5 & 6) | Huaxia Group, Sorting Hat, Ministry of Magic, Diagon Alley; Sirius Black stays Sirius Black without translating the surname | ID: Grup Huaxia; ES: Grupo Huaxia; TH: กลุ่มหัวเซี่ย; VI translates descriptive Western terms (Nón Phân Loại, Bộ Pháp thuật, Hẻm Xéo) and keeps Sirius Black |
When a novel goes deep into genre-specific terminology, Step 5 works alongside NovelVerba's built-in genre packs. You can explore how cultivation realms and alchemy terms map into English in Xianxia cultivation realms and terms in translation: from Qi Refining to Tribulation Crossing without literal traps, how Dou Qi ranks and magic classes are structured in Xuanhuan and western fantasy terms in translation: Dou Qi ranks, technique tiers, and magic classes, how status screens, dungeons, and cheats are handled in System and LitRPG web novel terms in translation: status screens, dungeons, quests, and cheats, how modern corporate, showbiz, and antique terms work in Chinese urban novel terms in translation: showbiz, antique appraisal, and business, and how Japanese name suffixes and pronouns are handled in Japanese honorifics and pronouns in Light Novels: -san, senpai, Ore, and Boku.
How do I lock names in the "Names & terms" tab (Step 1) and what are the 4 system limits I should know?
Knowing the boundary between default translation rules (Steps 2 through 7) and your active dictionary (Step 1) helps you avoid surprises across a 500-chapter book. First, Steps 2 through 7 are default system instructions that guide the AI when a term is not yet in your glossary, whereas only Step 1 is a hard lock resolved in strict 3-tier priority: the Names & terms tab (for that specific novel) > My dictionary (your personal cross-novel entries) > Genre packs. Because AI output can still drift across long chapter ranges, any recurring character, place, or technique should be locked in the Names & terms tab. Remember too that adding or editing a dictionary entry applies only to chapters translated afterward; it does not retroactively rewrite already-translated chapters unless you use line retranslation or retranslate those chapters.
Second, Step 2 (restoring CJK-transcribed foreign names like 赫敏 -> Hermione or 哈利·波特 -> Harry Potter) only triggers when the original name is confidently identifiable. For obscure Western names or author-invented fantasy names written in Chinese characters, Step 7 instructs the AI to fall back to conservative phonetic Pinyin, though it can still guess wrong on rare names until you scan the opening chapters and lock the intended spelling in Names & terms. Third, regarding source-language coverage and per-segment glossary size: Step 3 defines general rules for Chinese and Japanese personal names while Korean surname-first order is explicitly noted in the Spanish target rules; and when translating each text segment, the system activates only glossary entries whose source term actually appears in that segment, capped at a maximum of 12,000 glossary characters per segment.
Fourth, when you run the Automatic names & terms scan (up to 10 chapters per scan), candidates are grouped into 4 actions: Keep as-is, Translate meaning, Normalize, and Needs review. Confident proposals (Keep as-is, Translate meaning, Normalize, even if seen only x1) are pre-ticked, while uncertain Step 7 candidates (Needs review) and entries already in your dictionary remain unticked. Clicking Apply is always a human review gate: the system never overwrites your glossary without your confirmation. Locking your proper nouns first also ensures book titles and chapter headings match the body text, as detailed in Translating web novel book and chapter titles: keeping names aligned with the table of contents, before you export your finished book following How to translate an EPUB novel into English and export a file you can read.
- When creating a novel, select the matching Genre (Xianxia, Xuanhuan, Urban, Romance, or Web Novel) and make sure that pack is enabled under Genre packs in Shared dictionary.
- Go to the novel's Names & terms tab and run Scan names & terms across the first 5 to 10 chapters (up to 10 chapters per scan).
- Review the 4 proposal groups (Keep as-is, Translate meaning, Normalize, and Needs review): edit any rare foreign names or custom place renderings before clicking Apply.
- Translate subsequent chapter batches with your Step 1 Names & terms table locked so character names, sects, and locations stay consistent throughout the book.