amharic-corrector 1.0.4

High-performance Amharic phonetic query corrector

Amharic Corrector

High-performance phonetic query correction library for Amharic.

Changelog

1.0.4

• Single-word correction now uses phonetic skeleton similarity (consonant LCS) instead of frequency to rank candidates — produces phonetically correct results regardless of corpus frequency bias.
• Removed Levenshtein from the inner skeleton index loop (was causing ~35% speed regression); kept it only in the transliteration scoring path where it adds precision.
• Multi-word correction unchanged: co-occurrence / bigram context still drives ranking as intended.

1.0.3

• Internal release.

1.0.2

• Added instance-level caching to AmharicQueryCorrector to prevent shared state when multiple instances are used with different corpuses.
• Minor performance improvements in candidate lookup.

Installation

pip install .

Usage

from amharic_corrector import AmharicQueryCorrector
c = AmharicQueryCorrector("path/to/corpus.json")
print(c.correct_phrase("mekanitu"))

CLI

amharic-correct "Kuratie" --model path/to/corpus.json