This is reimplementation of the Amharic Metaphone algorithm of the Text-TransMetaphone package. This implementation uses an object oriented interface and will generate keys in Ethiopic script by default. IPA keys remain available and may be set at instantiation time or after wards. See the POD for details. The script 'examples/matchtest.pl' will perform matching on a group of 116 correctly spelled words compared against 166 misspelled words in 9 misspelling categories. In a few cases the misspelled words go outside the scope of what the Metaphone algorithm is intended for. The Text::Metaphone::Amharic algorithm does produce better results than the alternatives. What alternatives? :-) The only somewhat sensible comparison would be to transliterate Amharic into an ASCII system and apply the DoubleMetaphone to the converted words. Results for 4 transliteration systems are as follows ("low" granularity setting): Text::Metaphone::Amharic (Ethiopic) - 116 canonical words matched 160 of 166 error words. Text::DoubleMetaphone (SERA) - 116 canonical words matched 105 of 166 error words. Text::DoubleMetaphone (Ethiop) - 116 canonical words matched 99 of 166 error words. Text::DoubleMetaphone (ISO) - 116 canonical words matched 97 of 166 error words. Text::DoubleMetaphone (Mainz) - 116 canonical words matched 89 of 166 error words.