Modifier

Partager via

Text boundaries

  • Article

Applications like Microsoft Word provide several features that require identifying text boundaries of the supported writing systems. For example:

  • You can view document information like the number of characters, words, lines, or paragraphs.
  • If you press Ctrl + right arrow, your cursor moves to the next word.
  • If the text that you're entering extends past the margin, the text wraps to the next line.
  • If you press F8 repeatedly, the selection expands to the entire word, sentence, paragraph, section, and document.

Character boundaries

Unicode supports features like combining characters and ligatures. When determining a count of characters, each character might comprise one or more Unicode code points. As discussed in Text layout, the words “Tiếng Việt” (“Vietnamese” in Vietnamese) contains the character ệ, which could comprise one of four combinations of Unicode code points. However, you expect that “Việt” contains 4 “characters”. Counting the number of characters in a document requires an understanding of where one character ends and where the next character starts.

Word boundaries

It's difficult to have a concise definition of the term “word” that applies to all languages, writing systems, and scripts; however, we can consider a "word" to be a sequence of characters that have a specific meaning. Words are composed of one or more morphemes and represent combinations of syllables. Writing systems like Latin and Arabic can use spaces to separate words. Vietnamese uses the Latin script, but spaces separate syllables rather than words. Writing systems used for other East Asian languages, such as Thai, Japanese, and Chinese, don't separate words or syllables with word separators like spaces.

Features like selecting the word or moving to the next word in the document require an understanding of word boundaries in the various writing systems. Rules for implementing word boundary identification might depend on the script used, dictionary lookup, or heuristics. In addition, within one language, differences in conventions or between style guides might require alternate interpretations of how to define word boundaries. For more information, see Line and word breaking.

Line-break boundaries

When displaying text that spans multiple lines, separate lines are a result of:

  • a line breaking character (for example, line feed U+000A)
  • text that has been automatically wrapped to fit within a predetermined boundary (for example, page margins in a word processor)

Various algorithms exist to calculate the most appropriate line break location. Location for line breaks could include word boundaries, before or after specific characters, or be enabled using automatic hyphenation tools. For more information, see Line and word breaking.

Sentence boundaries

In English, several characters can indicate the end of a sentence, including the question mark, the exclamation mark, and the period. However, a period can also be used to indicate an abbreviation, as a decimal separator, and as a delimiter (for example, web address). Like line-breaking, semantic analysis is often needed to determine whether a punctuation mark indicates the end of a sentence or is being used for a different purpose.

Using locale-aware and language-aware methods for boundary detection

Both .NET and Java have classes that facilitate boundary detection. For Windows, you can use the SelectableWordsSegmenter class, while Java and ICU provide the BreakIterator class.