Can machines translate film? Artificial intelligence and film translation

Abstract

I have investigated how artificial intelligence (AI) operates in case of film translation. The
research was based on the dialogue list of three movies, in Spanish-Hungarian language pair.
Human speech can only be translated in context. In movies this context is created by
cinematographic instruments. In movies the most important context of the spoken words, is the
image itself. For an AI-based translator, every language is a ‘foreign language', every context
is a ‘foreign context', as it doesn't understand any, yet it translates from many languages into
many languages. Even if AI takes into account certain contexts, these are only related to the
linguistic environment and have nothing to do with the image. From the point of view of
applicability, I came to the following conclusion: as the requirements for AI-based and human
translation are the same (equivalence, adequacy, fidelity), then the currently used machine
translation is not suitable for movie translations. In order to evaluate machine-based subtitling
and machine-based translation of film scripts, a change of approach is required. The common
categories of translators have excessive expectations, though in most cases there is no need for
this, as the visual context helps a lot in comprehension. So, if we would like to get just basic
information about the plot, and the goal is to be able to follow a movie, the use of machine
translation is still suitable for this purpose.