Changelog

Changed initialism detection to add conservative undotted mode and diagnostic capabilities

Changed undotted initialism handling to use separator-aware run detection and vowel-preserving fallback

Updated changelog entries and release metadata for v0.2.8

Added reviewed source-spelling entries for common English and technical initialisms

Changed reviewed initialism precedence and provenance while retaining conservative unknown-uppercase behavior

Hardened conservative undotted initialism handling and expanded reviewed locale coverage

Documented reviewed initialisms, lexical-acronym exclusions, and downstream structured-normalization ownership

Added distilled regression coverage for initialism safety and provenance

Added opt-in policy controls for spelling bounded undotted uppercase initialisms with source, upper, or lower casing

Added reviewed source-spelling entries for common English and technical initialisms

Changed ambiguous one-letter units to require separators for identifier-like compact forms

Changed ambiguous English dotted forms to prefer letter readings without positive place evidence

Changed reviewed initialism precedence and provenance while retaining conservative unknown-uppercase behavior

Fixed reviewed German organization initialisms to render with source graphemes such as G m b H and A G

Fixed compass and geographic expansions that corrupted names and initials

Documented abbreviation-stage ownership and the minimum released dependency handoff for Spokenform

Documented opt-in initialism orchestration and structured-code ownership boundaries

Documented reviewed initialisms, lexical-acronym exclusions, and downstream structured-normalization ownership

Added unit decision diagnostics and PolyNorm regression coverage for compact-code precision

Added focused and benchmark-derived regression coverage for abbreviation ambiguity and source-aligned punctuation

Added distilled regression coverage for initialism safety and provenance

Added conservative multilingual abbreviation coverage, punctuation aliases, and longest-match German compounds

Added bounded Italian and German context selection and protected English prepositional in from inch expansion

Fixed English and German lexical expansions and protected dotted fragments in hyphenated initials

Documented the v0.2.6 release handoff and Spokenform dependency baseline

Improved boundary, context, unit, mapping, registry, and regression coverage for conservative lexical normalization

Improved street address and temperature abbreviation expansion

Fixed version fallback behavior

Added Spanish language support with unit abbreviations and quantity matches

Added Italian language support with unit abbreviations and quantity matches

Added Portuguese language entries for unit abbreviations

Added Czech language support with unit abbreviations and quantity matches

Improved English unit abbreviations with additional quantity matches

Updated language source documentation and API reference

Added French euro, US dollar, and pound sterling identities for prefix and suffix quantities

Added exact-span French dotted min. and sec. duration aliases while preserving standalone lexical min. behavior

Changed dotted quantity expansion to preserve sentence-final punctuation

Documented the boundary between identity recognition and downstream French numeric and currency grammar

Added structured quantity matches with exact spans and stable canonical identities

Added German electrical, frequency, magnitude, count, litre, and EUR symbols

Changed numeric magnitude priority while retaining standalone German abbreviations

Documented the boundary between quantity-symbol recognition and downstream number and grammar realization

Added explicit protected spans, source-aligned expansion traces, truthful instance-local unit customization, eager registration validation, registry audits, and dense-input regression coverage

Added immutable source-aligned replacement results, reviewed German compound aliases, and deliberate registry snapshot metadata for those aliases

Added Dutch, Polish, Russian, Swedish, and Turkish language support

Improved spoken form API with context-aware abbreviation expansion

Corrected punctuation-safe abbreviation boundaries, ellipsis-adjacent matching, immediate guard enforcement, and deterministic custom overrides

Corrected anchored followed-by guards after preceding source text

Made unsupported compound-looking unit expressions fail closed and added the Greek small-letter mu unit alias

Improved Unicode-aware German title and English address context decisions

Documented that unit expansion is lexical symbol expansion and does not own quantity verbalization or grammatical plural selection

Added API reference, customization guide, and language source documentation

Added edge case robustness tests and language-specific test coverage

Added spaCy POS vector support for improved abbreviation detection accuracy

Improved POS vector handling to reduce false positives

Fixed German example script

Added example scripts for abbreviation expansion and multi-language demos

Added numeric-aware unit expansion across 7 languages

Improved English abbreviations with direction guards and embedded-initialism fix