The Evolutive Nature of Translation and Teaching Workflow: Textometric Analysis of Multiple Revision Cycles in a Simulated Website Translation Project

Abstract

In this paper, we explore the practical and theoretical status of revision in the context of a Masters-level website translation project integrating machine translation (MT) for teaching purposes. We use textometric analysis to explore the nature of modifications made at each of the different stages of revision during the lifetime of one particular project. Computation of characteristic elements and Correspondence Analysis (CA) reveal regular patterns of linguistic intervention, with significant differences observed between the initial stages of the project (MT-EN0 generated by SYSTRAN Pure Neural Machine Translation and post-edited MT-EN1) and subsequent revision cycles. We combine quantitative and qualitative research methods to study characteristic revision patterns in different project cycles (chronological textual series). While some of these patterns represent unpredictable “punctual” edits, the majority are “progressive”, involving modifications that are predictable or gradually built up throughout the project. More generally, it appears that different types of revision closely mirror the guidelines that are presented to students during the teaching workflow. This observation leads us to posit an “evolutive” view of the translation project: just as each webpage in the project has to undergo several cycles of revision, so each stage of revision needs to be carefully primed during the preparation of the project with the clients as well as the students.