Slash-Dash

A Node.js package that removes em-dashes and en-dashes from text and replaces them with appropriate punctuation using natural language understanding.

Try it out: Playground

Installation

npm install slash-dash

Usage

import { processDash } from 'slash-dash';

const text = "The weather was perfect—sunny and warm—for our picnic.";

// Process both em-dashes (—) and en-dashes (–) - default behavior
const { processedText, meta } = await processDash(text);

console.log(processedText);
// "The weather was perfect, sunny and warm, for our picnic."

// Process only em-dashes
const { processedText: emOnly } = await processDash(text, { dashTypes: ['em'] });

// Process only en-dashes
const { processedText: enOnly } = await processDash(text, { dashTypes: ['en'] });

console.log(meta);
// [
//   {
//     index: 24,
//     replacement: ", ",
//     confidence: 0.5,
//     reason: "default_fallback",
//     contextSnippet: "..."
//   },
//   ...
// ]

Methods

processDash(text, options)

Processes text to replace em-dashes and/or en-dashes with appropriate punctuation.

Parameters:

• text (string): The text to process
• options (object, optional): Configuration options
  • dashTypes (string[], optional): Types of dashes to replace. Defaults to ['em', 'en']
    • 'em': Replace em-dashes (—, ---, --)
    • 'en': Replace en-dashes (–)

Returns:

• Promise<{ processedText: string, meta: Array }>
  • processedText: The text with dashes replaced
  • meta: Array of metadata objects for each replacement, containing:
    • index: Original position of the dash
    • replacement: The punctuation used as replacement
    • confidence: Confidence score (0-1)
    • reason: Reason for the replacement
    • contextSnippet: Context around the replacement

How It Works

The package uses compromise (a JavaScript NLP library) for natural language understanding to determine the most appropriate punctuation replacement for each em-dash or en-dash.

The NLP-powered algorithm analyzes:

• Parts of speech - identifies nouns, verbs, pronouns, determiners, and conjunctions
• Sentence structure - detects sentence boundaries and independent clauses
• Grammatical patterns - identifies appositives, explanations, relative clauses, and lists
• Context clues - examines surrounding words, punctuation, and grammatical relationships
• Clause types - recognizes independent and dependent clauses using subject-verb detection

Additional heuristics keep the output stylistically natural:

• Mirrored clauses (e.g. “Windows rattled—hearts did too.”) preserve the dash rhythm
• Intro adverbs & short lead-ins (e.g. “Sometimes, …”) become commas for smoother flow
• Copular explanations (e.g. “It was a quick trip—a weekend getaway.”) elevate to colons
• Punchy fragments (e.g. “—not this time.”, “—again.”) retain the dash emphasis

Possible replacements:

• Commas (,) - for most general cases, lists, continuations, and dialogue
• Periods (.) - for sentence boundaries or complete thoughts
• Colons (:) - for introducing explanations, definitions, or appositives
• Semicolons (;) - for connecting related independent clauses
• Spaces - for purely decorative em-dashes

Testing

npm test

Requirements

• Node.js >= 14.0.0

License

MIT