Abstract
Language availability, translation quality, and interface usability are distinct outcomes, yet comparisons of widely supported and underrepresented languages can collapse them into a single ranking. This paper critically reviews eleven selected sources relevant to Amharic, Afaan Oromo, low-resource translation, dataset quality, and human evaluation. The synthesis finds direct research on both languages, evidence of language-specific resource construction, and reasons to question the transfer of prose benchmarks to interface tasks. It also identifies reporting inconsistencies that make unqualified numerical rankings inappropriate for one examined English-Amharic study. The paper contributes a coverage-and-quality decomposition, an account of cross-language comparability, and criteria for interpreting a software-domain language gap. No current provider-support audit, new translation benchmark, or estimate of a gap against European languages is reported. The central conclusion is that an interpretable language-gap claim requires aligned tasks and transparent coverage, together with qualified language-specific evaluation.
- 0
- Views
- 0
- Downloads
- —
- Citations
- 0
- Cited here
not yet indexed
papers on ScholarXIV
- Subjects
- cs.AI
- Keywords
- Amharic · Afaan Oromo · low-resource translation · benchmark validity · software localization
- AI authorship disclosure
- Human Written
- Licence
- CC BY 4.0
- Submitted as
- PDF only
Submission history
-
First posted.