Evaluating Translation Quality in the ELSST Multilingual Thesaurus

The European Language Social Science Thesaurus (ELSST) was designed to promote multilingual subject access to the data holdings of the Consortium of European Social Science Data Archives (CESSDA). Currently available in 13 languages, its efficacy depends largely on the quality of its translations, all of which are derived from the English source.

This presentation describes work recently undertaken as part of the EU-funded Synergies for Europe's Research Infrastructure in the Social Sciences (SERISS) project to evaluate the translation quality of ELSST. Two evaluation methods were compared and contrasted: back-translation of the translated terms into English, and comparison of the set of ELSST terms assigned in different languages to the same cross-national studies. A set of questions were specially indexed for the second experiment.

Results of both methods were analysed to see if they could uncover differences in interpretation of the source terms and their translations, and were used to improve the ELSST translation guidelines, and to inform ongoing translation work. They will be of interest both to developers of multilingual vocabularies and to the indexer community, since they also contribute to the ongoing debate on how, and where in the data lifecycle, questions should be indexed.