Wednesday, February 12, 2014

The Gallup World Poll and African languages

Gallup, Inc. has conducted its World Poll survey in 160 countries annually since 2005. Such an undertaking naturally involves many translations of the questionnaires (research instruments) for use in many different societies. According to Gallup's Country Data Set Details (2005-2013), in the 26 African countries where the poll was taken, questionnaires were translated into the following African languages, in addition to the Europhone official languages of most countries - English, French, or Portuguese (this list was keyed in from the country-by-country info in their data table; outlinks on language names are to Wikipedia articles):

This represents a significant effort, involving translations that are, as I understand it, done on the country level. It also implies a number of issues that would be interesting to explore further, such as how translations are handled for the several cross-border languages used in the poll (indicated in the above list by two or more country names in parentheses). A related issue is whether there is any cross-reference in the translation process between versions in very closely related languages (such as Kinyarwanda and Kirundi, or the Manding languages,.Bambara and Dioula (Jula), or the Nguni languages, Ndebele, Xhosa, and Zulu).

Another issue is assuring equivalency among translations from the original language (English) version into the whole range of languages. This is an obvious challenge for something of such wide multinational scope, for which the usual equivalency testing would be too hard and expensive (at least with available human language technology). How well can tight revisions of translations from the main language version(s) country-by-country, pair-by-pair assure that the same thing is being asked across diverse languages?

Yet another issue is how easily enumerators who speak the languages fluently, but have had little or no familiarity with the written forms of those languages can become competent in reading and administering questionnaires in them. Most if not all of the above languages have a written form, but some of these have had limited use - schools in Africa often focus uniquely on Europhone languages, and it's not uncommon to hear educated Africans say that they cannot read or write their mother tongue.

Finally, thinking about survey work in African languages generally, what might be the possibility of survey companies like Gallup and scholars doing academic research somehow sharing data on their translated questionnaires? Could this be useful towards developing terminologies for all future survey work, as well as for any attempts at analysis in languages of the surveyed?

No comments: