Keyboard Support

Contact and Search

Keyman.com Homepage

Header bottom

Keyman.com

Other versions
Version 18.0 (home page)Version 17.0Version 16.0 (current version)Version 15.0Version 14.0Version 13.0Version 12.0 (home page)Version 11.0 (home page)Version 10.0 (home page)Version 9.0 (home page)Version 8.0 (home page)Version 7.0 (home page)Version 6.0 (home page)Version 5.0 (home page)Version 4.0 (home page)

On this page

Word breaker

The trie family of lexical models needs to know what a word is in running text. In languages using the Latin script—like, English, French, and SENĆOŦEN—finding words is easy. Words are separated by spaces or punctuation. The actual rules for where to find words can get quite tricky to describe, but Keyman implements the Unicode Standard Annex #29 §4.1 Default Word Boundary Specification which works well for most languages.

However, in languages written in other scripts—especially East Asian scripts like Chinese, Japanese, Khmer, Lao, and Thai—there are is no obvious break in between words. For these languages, there must be special rules for determining when words start and stop. This is what a word breaker function is responsible for. It is a little bit of code that looks at some text to determine where the words are.

You can customize the word breaker in three ways:

Overriding script defaults

The default word breaker makes assumptions about how each script (alphabet, syllabary, or writing system) works. You can override the defaults by specifying the overrideScriptDefaults option.

There is currently only one override:

'break-words-at-spaces'
Only breaks words at spaces for scripts that otherwise do not use spaces in between words.

Break words at spaces

This applies only to languages that borrow the Burmese, Khmer, Lao, or Thai scripts. The majority languages for these scripts do not use spaces in between words; hence, the default word breaker will produce undesired results when breaking words in these scripts. However, if your language is written in one of these scripts and does use spaces in between words, then you can set overrideScriptDefaults: 'break-words-at-spaces', to ensure word breaks do not occur in the middle of words, but instead, at spaces.

Your model definition file should look like this:

const source: LexicalModelSource = {
  format: 'trie-1.0',
  sources: ['wordlist.tsv'],
  wordBreaker: {
    use: 'default',  // we want to use the default word breaker, BUT!
    // Override the default for Burmese, Khmer, Lao, or Thai:
    overrideScriptDefaults: 'break-words-at-spaces',
  }
};

export default source;

Customize joining rules

The default word breaker is very liberal in what it considers is a word.

For instance, the default word breaker will split words at hyphens. Consider the following Plains Cree example; this is a single word:

amiskwaciy-wâskahikan

However, the default word breaker will produce three words: amiskwaciy, -, and wâskahikan.

To join words at hyphens and any other punctuation, provide the joinWordsAt option in the model definition file:

const source: LexicalModelSource = {
  format: 'trie-1.0',
  sources: ['wordlist.tsv'],
  wordBreaker: {
    use: 'default',     // we want to use the default word breaker, BUT!
    // CUSTOMIZE THIS:
    joinWordsAt: ['-'], // join words that contain hyphens
  }
};

export default source;

You can specify one or more strings to join words at:

const source: LexicalModelSource = {
  format: 'trie-1.0',
  sources: ['wordlist.tsv'],
  wordBreaker: {
    use: 'default',
    // CUSTOMIZE THIS:
    joinWordsAt: ['-', ':', '@'], // join words at hyphens, colons, at-signs
  }
};

export default source;

Writing a custom word breaker function

The word breaker function can be specified in the model definition file as follows:

const source: LexicalModelSource = {
  format: 'trie-1.0',
  sources: ['wordlist.tsv'],
  // CUSTOMIZE THIS:
  wordBreaker: {
    use: function(text: string): Span[] {
      // Return zero or more **spans** of text:
      return [];
    },
  },
  // other customizations go here:
};

export default source;

The function must return zero or more Span objects. The spans, representing an indivisible span of text, must be in ascending order of their start point, and they must be non-overlapping.

A Span object

A span is an indivisible piece of a sentence. This is typically a word, but it can also be a series of spaces, an emoji, or a series of punctuation characters. A span that looks like a word is treated like a word in the trie-1.0 model.

A span has the following properties:

{
  start: number;
  end: number;
  length: number;
  text: string;
}

The start and end properties are indices into the original string at which the span begins, and the index at which the next span begins.

length is end - start.

text is the actual text of the string contained within the span.

Example for English

Here is a full example of word breaker function that returns an array of spans in an ASCII (English) string. Note: this is just an example—please use the default word breaker for English text!

const source: LexicalModelSource = {
  format: 'trie-1.0',
  sources: ['wordlist.tsv'],
  // EXAMPLE BEGINS HERE:
  wordBreaker: function(text: string): Span[] {
    // A span derived from a JavaScript RegExp match array:
    class RegExpDerivedSpan implements Span {
      readonly text: string;
      readonly start: number;

      constructor(text: string, start: number) {
        this.text = text;
        this.start = start;
      }

      get length(): number {
        return this.text.length;
      }

      get end(): number {
        return this.start + this.text.length;
      }
    }

    let matchWord = /[A-Za-z0-9']+/g;
    let words: Span[] = [];
    let match: RegExpExecArray;
    while ((match = matchWord.exec(phrase)) !== null) {
      words.push(new RegExpDerivedSpan(match[0], match.index));
    }

    return words;
  },
  // other customizations go here:
};

export default source;

See also

The TypeScript definition of WordBreakingFunction and Span


Return to “Advanced Lexical Model Topics”