Skip to content

1.2.2

Choose a tag to compare

@max-fluff max-fluff released this 23 Jul 15:58
· 22 commits to main since this release

Added

  • Broader morphology across the built-in languages (Matching). Many more inflected forms in your text now match a heading:
    • English — classical and irregular plurals (indicesindex, criteriacriterion, analysesanalysis), native irregulars (micemouse, feetfoot), -f/-ves (wolveswolf), compounds (firemenfireman, salespeoplesalesperson) and Greek -sis plurals (synthesessynthesis).
    • German — finite verb forms (macht, machtemachen) and feminine -innen plurals (FreundinnenFreundin), through the CISTEM stemmer added beside the existing light one.
    • Russian — adjective case forms (новым, новыхновый), short nouns with a dropped vowel (днядень, снасон), suppletive stems reached from any case form (человекалюди), and -ята young-animal plurals (котятакотёнок).
    • Ukrainian — the same short-noun and fleeting-vowel coverage as Russian.
    • French — the closed -ail/-aux group and suppletive plurals (travauxtravail, yeuxœil).
    • Greek — the commonest modern plural classes (χώρεςχώρα, βιβλίαβιβλίο) and neuters (παιδιάπαιδί).

Changed

  • English now uses the Porter2 stemmer (Matching) in place of the original 1980 Porter algorithm. It over-stems less, so fewer unrelated words are drawn together — general and generous, universe and university, intern and internal now stay apart, while ordinary forms still link. Existing vaults may see a few English words start or stop linking; the index rebuilds itself.

Fixed

  • Light ending-strip mode no longer eats a silent e: notesnote, filesfile, valuesvalue now link, as do Spanish grandesgrande and French chevauxcheval.
  • Two files whose names differ only in case (Guide and guide) each open the file the row names, at its heading. Before, both rows in the Which one? list opened whichever the link resolver ranked first, so one file was unreachable.

Internal

  • The Porter2 port is checked letter-for-letter against the Snowball reference vocabulary (~42,000 words). CISTEM (MIT) and the light stemmers keep their upstream attribution and licence notices.