parse-latin

A Latin script language parser producing NLCST nodes.

• For semantics of nodes, see NLCST;
• For a pluggable system to analyze and manipulate language, see retext.

Whether Old-English (“þā gewearþ þǣm hlāforde and þǣm hȳrigmannum wiþ ānum penninge”), Icelandic (“Hvað er að frétta”), French (“Où sont les toilettes?”), parse-latin does a good job at tokenizing it.

Note also that parse-latin does a decent job at tokenizing Latin-like scripts, Cyrillic (“Добро пожаловать!”), Georgian (“როგორა ხარ?”), Armenian (“Շատ հաճելի է”), and such.

Installation

npm:

$ npm install parse-latin

Component.js:

$ component install wooorm/parse-latin

Bower:

$ bower install parse-latin

Duo:

var ParseLatin = require('wooorm/parse-latin');

Browser globals (both parse-latin.js and parse-latin.min.js are available):

<script src="path/to/parse-latin.js"></script>
<script>
  var parser = new ParseLatin();
</script>

Usage

var ParseLatin = require('parse-latin'),
    latin = new ParseLatin();

latin.parse('A simple sentence.');
/**
 * Logs something like:
 * ˅ Object
 *    ˃ children: Array[1]
 *      type: "RootNode"
 *    ˃ __proto__: Object
 */

latin.parse(
    'The \xC5 symbol invented by A. J. A\u030Angstro\u0308m ' +
    '(1814, Lo\u0308gdo\u0308, \u2013 1874) denotes the ' +
    'length 10\u207B\xB9\u2070 m.'
);
/**
 * Logs something like:
 * ˅ Object
 *    ˃ children: Array[1]
 *      type: "RootNode"
 *    ˃ __proto__: Object
 */

API

• ParseLatin(options?)
• ParseLatin#tokenize(value)
• ParseLatin#parse(value)

ParseLatin(options?)

Exposes the functionality needed to tokenize natural Latin-script languages into a syntax tree.

Parameters:

• options (Object, optional)

  • position (boolean, default: false) - Whether to add positional information to nodes.

ParseLatin#tokenize(value)

Tokenize natural Latin-script language into letter and numbers (words), white space, and everything else (punctuation).

ParseLatin#parse(value)

Tokenize natural Latin-script languages into an NLCST syntax tree.

var ParseLatin = require('parse-latin'),
    latin = new ParseLatin();

latin.parse('A simple sentence.');
/**
 * Object
 * ├─ type: "RootNode"
 * └─ children: Array[1]
 *     └─ 0: Object
 *           ├─ type: "ParagraphNode"
 *           └─ children: Array[1]
 *              └─ 0: Object
 *                    ├─ type: "SentenceNode"
 *                    └─ children: Array[6]
 *                       ├─ 0: Object
 *                       |     ├─ type: "WordNode"
 *                       |     └─ children: Array[1]
 *                       |        └─ 0: Object
 *                       |              ├─ type: "TextNode"
 *                       |              └─ value: "A"
 *                       ├─ 1: Object
 *                       |     ├─ type: "WhiteSpaceNode"
 *                       |     └─ value: " "
 *                       ├─ 2: Object
 *                       |     ├─ type: "WordNode"
 *                       |     └─ children: Array[1]
 *                       |        └─ 0: Object
 *                       |              ├─ type: "TextNode"
 *                       |              └─ value: "simple"
 *                       ├─ 3: Object
 *                       |     ├─ type: "WhiteSpaceNode"
 *                       |     └─ value: " "
 *                       ├─ 4: Object
 *                       |     ├─ type: "WordNode"
 *                       |     └─ children: Array[1]
 *                       |        └─ 0: Object
 *                       |              ├─ type: "TextNode"
 *                       |              └─ value: "sentence"
 *                       └─ 5: Object
 *                             ├─ type: "PunctuationNode"
 *                             └─ value: "."
 */

Syntax Tree Format

Note: The easiest way to see how parse-latin tokenizes and parses, is by using the online parser demo, which shows the syntax tree corresponding to the typed text.

Basically, parse-latin splits text into white space, word, and punctuation tokens. parse-latin starts out with a pretty easy definition, one that most other tokenizers use:

• A “word” is one or more letter or number characters;
• A “white space” is one or more white space characters;
• A “punctuation” is one or more of anything else;

Then, it manipulates and merges those tokens into an NLCST syntax tree, adding sentences and paragraphs where needed.

• Some punctuation marks are part of the word they occur in, e.g., non-profit, she\'s, G.I., 11:00, N/A, &c, nineteenth- and...;
• Some full-stops do not mark a sentence end, e.g., 1., e.g., id.;
• Although full-stops, question marks, and exclamation marks (sometimes) end a sentence, that end might not occur directly after the mark, e.g., .), .";
• And many more exceptions.

Benchmark

On a MacBook Air, parse-latin parses 2 large books, 27 big articles, or 2,485 paragraphs per second.

To put things into perspective, Shakespeare’s works contain 884,647 words. I have not tested it, but in theory parse-latin should parse these works in (slightly above) three and a half seconds.

             latin.parse(document);
  2,001 op/s » A paragraph (5 sentences, 100 words)
    235 op/s » A section (10 paragraphs)
     22 op/s » An article (10 sections)
      2 op/s » A (large) book (10 articles)

Related

• nlcst
• retext
• textom
• parse-dutch
• parse-english

License

MIT © Titus Wormer