Release Notes

February 10, 2026 · View on GitHub

Version 6.7-SNAPSHOT (In Development)

Overview

Java port of SymSpell v6.7.3 with all custom features preserved. This release brings upstream algorithmic improvements while maintaining backward compatibility.

What's New in v6.7

Improved Word Segmentation

  • Ligature Normalization: Automatic conversion of Unicode ligatures (fi → fi, fl → fl) using NFKC normalization
  • Hyphen Removal: Better handling of hyphenated words and syllabification artifacts
  • Case Preservation: Maintains uppercase on first character during word break segmentation
  • Punctuation Adjacency: Punctuation marks now stay adjacent to words (no extra spaces)

Safety Improvements

  • Added length checks before character access to prevent edge-case errors
  • Enhanced null safety in word segmentation

Custom Features (Preserved)

All 5 custom features remain fully functional:

  1. Exclusion Dictionary - Whitelist terms that bypass spell checking
  2. Edit Factor Configuration - Dynamic edit distance based on word length
  3. Ignore Unknown Words - Option to skip unknown terms
  4. Bigram Key Splitting - Configurable phrase-level corrections
  5. QwertzDistance - German keyboard layout support for better corrections

Technical Details

  • Performance: 9,523 queries/second (< 1% overhead from new features)
  • Accuracy: 88.61% maintained (Qwerty keyboard tests)
  • Tests: 59/59 passing
  • Compatibility: Drop-in replacement for v6.6

Migration Guide

From v6.6 → v6.7:

No code changes required! Simply update your Maven dependency:

<dependency>
    <groupId>io.github.mightguy</groupId>
    <artifactId>symspell-lib</artifactId>
    <version>6.7-SNAPSHOT</version>
</dependency>

Behavioral Changes:

  • Word segmentation now better handles uppercase text (e.g., "ITWAS" → "IT WAS")
  • Punctuation appears adjacent to words without spaces (e.g., "hello,world" → "hello, world")
  • Ligatures are automatically normalized before processing

All changes are non-breaking and only enhance existing functionality.

Code Changes

All changes are isolated to the wordBreakSegmentation() method in SymSpellCheck.java:

Input normalization (v6.7.0):

// Normalize ligatures and remove hyphens
phrase = java.text.Normalizer.normalize(phrase, java.text.Normalizer.Form.NFKC)
    .replace("-", "");

Case preservation (v6.7.0):

// Lookup with lowercase, preserve uppercase on first char
if (originalPart.length() > 0 && Character.isUpperCase(originalPart.charAt(0))) {
    resultChars[0] = Character.toUpperCase(resultChars[0]);
}

Punctuation adjacency (v6.7.0):

// No space before punctuation/apostrophes
if (isPunctuation || isApostrophe) {
    compositions[destinationIndex].setSegmentedString(
        compositions[circularIndex].getSegmentedString() + part);
}

Testing

Test Coverage:

  • Original SymSpell features: 20 tests
  • v6.7.x new features: 19 tests
  • Custom features: 19 tests
  • Accuracy/performance: 1 test
  • Total: 59 tests, all passing

Verified Against:

  • Original C# SymSpell v6.7.3 test data
  • Regression tests for all custom features
  • Performance benchmarks (no degradation)

Version History

v6.7 (Current)

  • Ported SymSpell v6.7.3 features
  • Enhanced word segmentation
  • All custom features preserved

v6.6

  • Bigram dictionary support
  • Sentence-level context for corrections

v6.5 and earlier

Last Updated: 2026-02-10 Status: Development Build: Passing (59/59 tests)