Writing without word gaps
In English and many other languages, spaces mark word boundaries, and software can break a line at any space. Not every writing system works this way. Chinese and Japanese are normally written without spaces between words. Thai uses spaces mainly to separate phrases or sentences, not individual words. Readers of these languages do not need word gaps; they recognize words from context and from the structure of the script.
For a subtitle, though, line breaking is a practical problem. A line has limited width, so long text has to be divided somewhere. If there are no spaces, where can the break go, and which breaks do readers find acceptable?
What word segmentation means
Word segmentation is the task of deciding where one word ends and the next begins in text that does not mark it. People do this without effort, but it is not trivial for software. Some character sequences can be divided in more than one way, with different meanings, and the correct reading depends on context. This is also an issue for other text tools. Search, spell-checking and automatic captioning engines all need some way of handling it.
Segmentation matters for subtitles in two ways. First, a good line break normally falls between words or phrases, not inside a word. Second, a break placed at an awkward point can change how a reader parses the sentence, just as it can in any language when a line break splits a phrase. Guidance on this in general is in the article on line breaks.
Break rules in Chinese and Japanese
Written Japanese and Chinese can be broken between most characters, since each character or short character group may form a unit. But conventions limit certain breaks. Japanese typesetting has long-established rules, often described as line-start and line-end prohibitions. In general:
- Certain closing marks, such as closing brackets and some punctuation, should not begin a line.
- Certain opening marks, such as opening brackets, should not end a line.
- Small characters, long-vowel marks and similar elements are usually kept attached to the preceding character.
- Marks that come in pairs should not be separated awkwardly.
The details differ between languages and between publishers, so a project should follow the rules that the commissioning party specifies. Chinese typesetting has its own, similar conventions, and the punctuation differs between regions.
Because Japanese mixes several kinds of symbol, an unbroken string of numbers or Latin letters inside a Japanese sentence is also treated as a unit that should not be divided arbitrarily.
Thai and similar scripts
In Thai, spaces generally mark the end of a phrase or sentence, so a line cannot be broken at a space in the way it can in English. Breaking inside words is considered poor style, and some character combinations must stay together because vowel and tone marks attach to the consonants around them. Software has to know the word boundaries to avoid splitting such groups, often by using a dictionary-based method. If the player does not do this, a line may break in the middle of a word, which native readers notice immediately.
Other languages that follow similar conventions exist, such as Lao, Khmer and Burmese, and the same general points apply. The specific rules should come from a reviewer who reads the language.
Character counts and line length
Subtitle guidelines often set a maximum number of characters per line. In Chinese, Japanese and Korean, the limits are typically lower than for alphabetic languages, because each character carries more meaning and is displayed wider. Exact limits depend on the client or broadcaster, so there is no universal figure to quote. Treat any number you find as a house style rather than a law, and use the project’s own guidelines. The wider topic is covered in line length.
Reading speed is also measured differently. Because a single character can carry what several English letters carry, a count of characters per second does not translate directly between languages. Style guides for each language set their own figures, and the principles in reading speed still apply: viewers must be able to read the text comfortably while keeping an eye on the picture.
Vertical text and ruby
Japanese has historically been written vertically as well as horizontally, but subtitles on screen are almost always horizontal. Some productions use small annotation characters placed alongside others to show pronunciation. These add height to a line and need extra spacing. Most subtitle formats do not support them well, so teams should check what the target player can display before planning to use them.
Automatic line breaking and its limits
Subtitle editing tools and video players use different methods for breaking lines. Some insert breaks only at spaces, which does nothing for spaceless text. Some break at any character, which can split words. Better tools use rules or dictionaries for segmentation. When a file is displayed in a player that uses its own line breaking, the result may differ from what the editor showed. As a precaution:
- Insert explicit line breaks in the subtitle file where possible, rather than relying on the player.
- Keep each line short enough to fit without a forced wrap.
- Preview in the actual players that viewers will use, such as web players.
- Check that the character encoding is correct so that characters are not damaged.
- Have a fluent reviewer read the result on screen.
Automatic tools can help produce a first draft of segmentation and timing, but the output needs a human check, since mistakes in this area can change meaning or look careless. Conventions vary by country, broadcaster and platform, so follow the style guide of the party that will publish the subtitles.