我一直在為這個問題頭痛，剛剛找到了一個解決方案，所以在此分享，以防其他人也發現自己試圖讓某些文字不被自動轉寫。這個解決方案既違反直覺又可疑。
簡而言之：
如果您有非口語文字，這些文字不應該被編碼轉換器自動轉寫，而是應該以其原始編碼發布，請將其包裹在自訂字元樣式中，並給該自訂樣式 nonpublishable 屬性。
是的，是 nonpublishable 屬性，而不是看起來更合適的 nonvernacular。nonvernacular 屬性實際上不會影響哪些內容被轉寫，而 nonpublishable 屬性目前不會影響哪些內容被發布，至少 Publishing Assistant 不會。
詳細背景：
我即將排版一個自動轉寫的專案。該專案包含某些國家語言的詞彙。這些可以在詞彙表條目、介紹性材料和腳註中找到，並在 \znp …\znp* 標記內標記，我在 custom.sty 中將其定義為 nonvernacular、publishable 字元樣式，因為事實就是如此。
在主要/來源專案中，口語文字和國家語言文字都在同一個書寫系統中。在自動轉寫的專案中，口語文字應轉換為僅用於該語言的傳統書寫系統，但國家語言的文字應以其原始書寫系統呈現。（傳統書寫系統無法表示國家語言中發現的卷舌音，而且無論如何，看到國家語言以口語語言的傳統書寫系統呈現都會很奇怪。）
但將此類文字標記為 nonvernacular 對防止其被轉寫沒有影響。
TECkit 映射表也不能被指示忽略 \znp …\znp* 標記內封閉的文字，因為標記本身不會被傳入映射表進行處理。（推測以前是這樣的，因為 Paratext 幫助文件仍然說，“請確保您使用的 TECkit.map 檔案保留 USFM 標記，以便它們不會被轉換。”）
棘手的解決方案是將 publishable 屬性更改為 nonpublishable。這並沒有阻止 Publishing Assistant 發布它。
我不知道這是否會與其他輸出路徑（SAB / Print Draft / PtxPrint / YouVersion via DBL）搞亂事情，無論是目前還是未來它們是否會悄悄丟棄不可發布的文字，所以這絕對是這個變通方案的一個顧慮。
除了 nonpublishable 使文字免於自動轉寫外，我不清楚這兩個屬性可能還有什麼其他影響。某處有文件解釋這一點嗎？
I’ve been beating my head against this problem, and just found a solution, so am sharing here in case anyone else finds themselves trying to keep certain text from getting auto-transliterated. This solution is both counter-intuitive and shady.
In a nutshell:
If you have non-vernacular text that should not be auto-transliterated by an encoding converter but rather be published in its original encoding, wrap it in a custom character style, and give that custom style the nonpublishable property.
Yes, the nonpublishable property rather than nonvernacular, as would have seemed fitting. The nonvernacular property does not actually affect what gets transliterated, and the nonpublishable property does not currently affect what gets published, at least not by Publishing Assistant.
Detailed background:
I’m about to typeset an auto-transliterated project. The project contains certain words in the national language. These can be found in glossary entries, in introductory material, and in footnotes, and are tagged within \znp …\znp* markers, which I’d defined in custom.sty as a nonvernacular, publishable character style, because that’s what it was.
In the primary/source project, the vernacular text and the national language text are both in the same writing system. In the auto-transliterated project, the vernacular text should be converted into a traditional script used only for this language, but the text in the national language should come through as it was in its original script. (The traditional script cannot represent the retroflex sounds found in the national language, and it would be weird to see the national language in the vernacular language’s traditional script in any case.)
But marking such text as nonvernacular had no effect on keeping it from being transliterated.
Nor could the TECkit map be told to ignore text enclosed in \znp …\znp* markers, because the markers themselves are not passed into the map for processing. (Presumably they used to, because the Paratext Help documentation still says, “Be sure the TECkit.map file you use preserves the USFM markers so they will not be converted.”)
The tricky solution was to change the publishable property to nonpublishable. This did not prevent Publishing Assistant from publishing it.
I don’t know whether that will mess things up with other output paths (SAB / Print Draft / PtxPrint / YouVersion via DBL), whether currently or in the future they will quietly drop unpublishable text, so that is definitely a concern with this workaround.
Other than nonpublishable exempting text from auto-transliteration, it’s not clear to me what other effects either of these properties might have. Is there documentation somewhere that explains this?
