MoVoC: Morphology-Aware Subword Construction for Ge'ez Script Languages

Subword-based tokenization methods often fail to preserve morphological boundaries, a limitation especially pronounced in low-resource, morphologically complex languages such as those written in the Ge'ez script. To address this, we present Morpheme-aware Subword Vocabulary Construction (MoVoC) and train MoVoC-Tok, a tokenizer that integrates supervised morphological analysis into the subword vocabulary. This hybrid segmentation approach combines morpheme-based and Byte Pair Encoding (BPE) tokens to preserve morphological integrity while maintaining lexical meaning. To tackle resource scarcity, we release morpheme-annotated data for four Ge'ez script languages (169,806 records) and a morphemeaware vocabulary for two of them. Scoring every comparison on words held out from the morpheme resource that constrains the MoVoC-Tok applied to that language, MoVoC-Tok has the highest MorphScore in three of four languages and the highest morpheme boundary precision in two of four, with its largest and most clearly resolved gains on Tigre, a language it was not trained on and to which it is applied cross-lingually. The machine translation results reported in the published version are withdrawn. Data, tokenizers and evaluation code are publicly available. 1

This is a corrected version of Teklehaymanot et al. (2025). The evaluation artifacts (checkpoints, predictions, scoring pipeline) behind the results in the published version were not recovered, and the published figures could not be reproduced from the released code and data. Following the advice of the EMNLP 2025 programme chairs, every result in this version is replaced by results reconstructed from the public repository. The method (Section 3) is unchanged. The main corrections are:

1 https://github.com/hailaykidu/MoVoC

• Examples (Tables 1 and 6). The released tokenizers do not produce the segmentations shown in the published Table 1 and Section 7, and the published translation example could not be reproduced. Both are replaced by actual tokenizer output.

• Setup. Training hyperparameters, evaluation data and vocabulary sizes are corrected to those recorded in the repository.