Translation using crowdwourced IEC
Use collaborative evolutionary computation to translate and edit texts of varying size and complexity (e.g., print ads, monographs) in a matter of hours and at a fraction of the cost of current human editorial services.The resulting text will represent a local optimum: the best possible revision or translation of a given text relative to the aesthetic preferences of the target audience.This method pics up on textual features that are fatally dependent on evolving, human evaluations and therefore represents one potential defense against the wholesale displacement of human translators by AI.
Suppose you're translating a French novel into English: we decompose the text into manageable fragments - i.e., fragments small enough to decrease the possibility of user fatigue during the editor's repeated evaluation of each individual variation of a given translated fragment. We then download the fragments onto a crowdsourcing market consisting of representatives of a target audience, professional translators/editors, or non-professionals from which we can nevertheless reliably extract professional-level translations via a more conventional optimizing process. We then obtain multiple iterations of each translated fragment to be evaluated by the editors.The preferred variants are retained and the less favorable variants are rejected. As multiple cycles of this process are pursued via the evolutionary algorithm, an 'optimal' variant is eventually seized upon.The fragments are reassembled in order to form the complete, translated version of the novel.
E-mail Gabi Lipede