Crowdsourcing, corpus use, and the search for translation naturalness

Abstract

This paper argues that corpus use in translation and the Facebook non-professional crowdsourcing model both aim to create more natural-sounding translations. A number of studies on corpus use support this hypothesis, but, to date, there have been no empirical studies on whether crowdsourcing translations produces texts that comply with the conventions users expect, consequently appearing more natural. After a theoretical discussion on how corpus use and Facebook crowdsourcing both intend to achieve more naturally sounding translations, the empirical study contrasts the crowdsourced Peninsular Spanish version of Facebook to original Spanish social networking sites. The methodology is based on a comparable corpus (Baker 1995) and compares all the interactive segments, such as navigation menus and dialog boxes, in this version of Facebook to a similar corpus extracted from the top 25 social networks locally produced in Spain. The contrastive analyses focus on verbal use and terminological conventions. The results confirm that the linguistic features examined in Facebook and produced through a crowdsourced non-professional model match those found in the corpus of non-translated networking sites.