Capstan: Linguistic Quality Control

Guide

Guide

We engineer equivalence

When a language expert is involved early in a project, the impact on quality is immense: machine translation projects may require that the authors learn controlled writing.

Comparative research calls for a translatability assessment of the draft source version. A workflow needs to be designed and tested. Good translation and adaptation notes need to be drafted. Linguists need to be briefed, and they should have access to a glossary, a style guide and a helpdesk.

All this is part of Linguistic Quality Assurance (LQA): a set of processes designed to ensure that multiple language versions of a document will meet the highest standards of linguistic, cultural and functional equivalence.

Is this what you need? Contact us now

cApStAn Guide: case study

Objective

For the fourth European Quality of Life Survey (EQLS), Eurofound and cApStAn set up a translatability assessment of new items in the EQLS master questionnaire just before the translation process was due to begin; cApStAn was also asked to identify existing items that might pose translation problems, and to draft item-by-item translation and adaptation notes.

The objective was to identify all potential pitfalls translators were likely to face, to propose translation notes and, if needed, to suggest alternative wording that would circumvent the issue without loss of meaning.

Methodology

A team of 6 linguists representing 3 language groups (Dutch and Swedish for Germanic; French and Italian for Romance; Czech and Polish for Slavonic) went through the exercise of producing dummy translations of the draft Master Questionnaire. Each time the linguists encountered a translation difficulty, they mapped it to one of cApStAn’s translatability categories and described the issue (in English). This feedback was then analysed and collated by an in-house team of senior linguists led by Elica Krajčeva, who directed a special effort at generalizable and recurring issues.

Results

The resulting consolidated translatability report highlighted a number of ambiguous formulations; some potential cultural issues; a larger number of known issues with known workarounds; unnecessarily complex wording or syntax; and unintentional inconsistencies. A webinar was organised to scrutinise the report with both Eurofound and the contractor in charge of implementing the survey. The translation and adaptation notes became part of the augmented final source questionnaire sent to the translators.

Objective

cApStAn Translate was asked to propose a translation approach for admission exams to the University of Luxembourg: from English into German and French for some items, from German into French for others and from French into German for a third lot. The objective was to obtain three sets of admission exams that would be deemed equivalent by domain experts so as not to put one target group at an advantage or a disadvantage. It was also important to document each step of the procedure, every adaptation made, and every decision taken.

Methodology

A double translation design was set up: for each language combination, two independent translators each prepared their own target version of the assessment. The translators were selected based on their experience both as teachers and as translators. A third translator, usually the most senior one, merged the two translations, taking the best elements from each, and then discussed the resulting version with the two translators. The reconciled version was submitted to domain experts before the final proofreading took place.

Results

The double translation design, combined with a special focus on linguistic and formal characteristics that were most likely to drive the psychometric properties of the test items, maximized cross-language comparability for these admission exams. The communalities revealed by the item analysis were high, and no students who failed the examination felt that this might be due to a higher level of difficulty in his or her language. The entire translation and adaptation history of each test item in each language was recorded in a centralised monitoring tool.