196) A term table fixes inconsistency
197) Wording problems in subtitles come in two kinds: "this word is translated wrongly" and "the same word is translated differently from one place to another". The first is a dictionary problem; the second is what a term table is for.
198) The second is both more common and harder to spot: `Thor` becomes "Thor" in episode one and "Donar" in episode seven, and each episode looks fine on its own while the series as a whole makes viewers think a new character appeared. A term table turns "this production calls him Thor" into an executable rule applied to the whole subtitle file at once.
199) When it is worth building one
200) Not every subtitle file needs a term table. The cases that justify one are usually these:
203) Keeping one table across a season
204) A term table is maintained incrementally: settle the names in episode one, then add a rule whenever a new term appears, and by the end of the season the table is complete. For the next season of the same series, start from it.
205) Write rules a little wider than strictly needed — once you have decided on "Thor", add "Donar -> Thor" as well, so someone else's subtitles can be brought in line in the same pass.
206) Always read the change list afterwards: it names each cue, each line and what became what. It is the only place where you can confirm that the change was right and not overdone, and any entry you dislike can be reverted with a click.
209) What not to hand to a term table
210) A term table changes words. These three things are outside its job, and should be:
211) Grammar and sentence structure — you cannot turn a word into a sentence; sentence-level work goes back to Subtitle translation or to a human.
212) Content cleaning — dropping whole lines such as watermarks, ads and SDH cues belongs to Subtitle cleaning; term table rules replace, they do not delete lines.
213) Timelines — a term table never touches time. For a global offset or a rate problem go to Timeline adjustment.
214) A few things easy to get wrong
215) Chinese still replaces substrings with whole-word matching ticked — that is normal. Chinese and Japanese do not separate words with spaces, so the tool cannot determine word boundaries and whole-word matching has no effect there; the `art` will not touch `start` protection only works for English.
216) Order changes the result. Rules run from top to bottom and text produced by one rule is scanned again by the rules below it. Replace A with B and then B with C and you end up with C.
217) To change only the translation, set the scope. In a bilingual file, pick "second line only"; entries with just one line stay untouched under either single-line setting — the tool does not guess which side is the original.
218) The term table is not saved. It is empty every time you open the page, so keep your own text file and pass it on.
219) Related tools and guides
220) For the full walkthrough see How to batch-replace terms in subtitles; if the content is dirty run Subtitle cleaning first; once the names are settled and you need other languages go to Subtitle translation; and run the result through Subtitle QC.