Aleph
Volume 11, Numéro 4, Pages 255-287
2024-07-25

Evaluating Parallel Corpora Based Machine Translation Performance From English Into Arabic

Authors : Chetioui Oussama .

Abstract

Abstract: This article aims at providing a practical quantitative analysis to evaluating parallel corpora based machine translation (PCBMT) performances using an English into Arabic directionality. This article’s proposal is to compare PCBMT performances of two different source texts in two different situations. First situational PCBMT performance evaluation embarks from the non-existence of an online Arabic parallel text to the source text. The second situational PCBMT performance evaluation however, uses another source text which humanely translated parallel text is available online for public use. Subsequently, a quantitative analysis is held to gauge quality in both performances in the light of their different situations. Upon this, the present study explores the extent to which PCBMT systems’ performance variates between situation 1 and situation 2, in terms of observing Arabic structural and contextual requirements to provide the ultimate output. The study shows that PCBMT performance from English into Arabic is quite divergent between the two situations, and reiterates the need to further investigate Arabic renditions of PCBMT performances in both directionalities, from and into Arabic.

Keywords

Machine translation evaluation ; Parallel corpora ; PCBMT performance ; Structurality ; Context