v0.6.3

🔧 Build & Infrastructure

• Golang version bumped in:

  • go-ossf-slsa3-publish.yml
  • go.yml

• Integrated upstream num2words and classifier libraries.

---

🌐 Language Support & Models

• Hebrew3:

  • New phonemizer model, homograph model, and packed dictionary.
  • Added learned patterns.

• English:

  • Introduced homograph model.

• German:

  • Added new dictionary.

• Minnan:

  • Added new 2-series models.

• Added new languages (unspecified which, beyond those above).

---

🧠 Core Features

• Integrated a sentencizer.

• Enabled loading of abbr.tsv into dictionaries.

• Finalized model for homograph result marking.

• Parallelized homograph test.

• Homograph-related improvements and documentation updates:

  • HOMOGRAPH.md, ROADMAP.md, README.md, etc.

---

🛠️ Other Changes

• Added zip model loader.
• Improved logging: now logs errors explicitly.
• Updated quaternary dependency to v0.2.0.
• Added and updated dev contribution docs: CONTRIBUTING.md, DEVELOPING.md.

Full Changelog: v0.6.2...v0.6.3

v0.6.2

• repacked all dictionaries for forward / reverse phonemization (new patch version)
• fix bug: Words with identical IPA as Text were vanishing
• New numToWords languages:
  • Czech
  • German
  • Spanish
  • French
  • Hungarian
  • Polish
  • Russian
  • Slovak
  • Ukrainian

Full Changelog: v0.6.1...v0.6.2

v0.6.1

• fix compilation on 386

Full Changelog: v0.6.0...v0.6.1

v0.6.0

• retrained all models for forward / reverse phonemization (breaking change, new minor version)

• new transformer based models for homograph word inference (english)

• working study initiation from null, {} map in language.json

• Handle numerics in Arabic and English

• Add groups of same diacritics to hebrew2

• analysis2 normalization/diacritics sorting

• Fixed -overwrite in homograph scripts

• Split on hanzi/nonhanzi boundary

• don't crash if language dict not found in phonemization steps

• User interface

  • toucan TTS integration
  • word editor

• New Languages:

  • cantonese
  • minnan/taiwanese
  • minnan/hokkien

What's Changed

• add build.sh to cmd by @thewh1teagle in #27
• rename dirty to lexicon by @thewh1teagle in #26
• format all go files by @thewh1teagle in #29
• standard format and dev recommendation in docs by @thewh1teagle in #30
• add batchsize flag and remove append flag from learn.tsv by @thewh1teagle in #36

Full Changelog: v0.5.1...v0.6.0

v0.5.1

• retrained 49 new models for forward / reverse phonemization (non-breaking change, new patch version)
• enable normalization for Vietnamese
• fix normalization non working
• split hyphen words
• remove old analysis aligner
• fix cleaning bug and do final clean
• backend: add explain world feature
• Starting with empty/null json.Map using short keywords
• fix: resolve infinite loop and improve alignment logic with Efficient Memoization for padspace languages
• Add new rule SrcDuplicate to language.json
• Clear combobox on first partial search click
• Don't overwrite model when training
• Finetune italian/tamil using --rowlossimportance 6

What's Changed

• update gitignore by @thewh1teagle in #17
• add hebrew2 folder by @thewh1teagle in #13

New Contributors

• @thewh1teagle made their first contribution in #17

Full Changelog: v0.5.0...v0.5.1

v0.5.0

• retrained all 87 models for forward / reverse phonemization (breaking change, new minor version)
• new transformer based models for out-of-dictionary word inference
• feature: punctuation preservation: toggle preserve/hide punctuation mode
• feature: multiple language dictionaries for use in the same sentence based on user preference

User interface

• out of dictionary words in underlined
• preserve punctuation in bold

New Languages:

• english/american
• english/british

Thanks to kokoro for generously offering their data.

Full Changelog: v0.4.0...v0.5.0

v0.4.0

• retrained all 85 models for forward / reverse phonemization (breaking change, new minor version)
• new quaternary models (should be faster)
• threading racing fix
• dictionary processing error fix

Full Changelog: v0.3.0...v0.4.0

v0.3.0

• retrained all 85 models for forward / reverse phonemization (breaking change, new minor version)
• fixes regarding thread safety of goruut

Full Changelog: v0.2.4...v0.3.0

v0.2.4

• fix compilation on architectures without avx512
• reverse phonemization models for all provided languages (from IPA to the language)
• fix forward direction phonemization to work (load the model)

Full Changelog: v0.2.3...v0.2.4

v0.2.3

• fix compilation on architectures without avx512
• reverse phonemization models for all provided languages (from IPA to the language)

Full Changelog: v0.2.2...v0.2.3