In spite of the advantages of professional translation, some companies rely on machine translation (MT) software to speed up the process. While the quality of machine translation has recently been improved, the results are still deficient.
To explore the limits of MT, you can perform some tests using the machine translation software you prefer. I asked the software to translate the set phrase, “Buscarle cinco patas al gato,” which is a Spanish expression that has a meaning similar to “Complicating things unnecessarily” in English. In this case, MT gives us the following result: “Look for five-legged cat,” a sentence that can definitely confuse users.
“Look for five-legged cat” is quite a literal translation. The machine does not know that cats with five legs do not really exist, which is not the point of the Spanish meaning. Spanish speakers have never seen a cat with five legs, they just search for it.
In general, the main disadvantage of MT is the fact that this kind of software isn’t capable of understanding the context of a sentence. Therefore, MT fails completely when a polysemic meaning or figurative language comes into play. These figures of speech require interpretation, and only a human translator can do this well.
We all live in the so-called immediacy society, and if there is anything that MT does well, it is to offer an immediate result. But we all know that a good translation, especially if it’s a literary one, requires time for research, thinking and answering questions about demanding sentences in order to arrive at the best equivalent meaning in the target language.
Let me give you another representative example. Now I will use the first paragraph of a literary masterpiece, El Aleph written by Borges in 1945. The original text is:
“La candente mañana de febrero en que Beatriz Viterbo murió, después de una imperiosa agonía que no se rebajó un solo instante ni al sentimentalismo ni al miedo, noté que las carteleras de fierro de la Plaza Constitución habían renovado no sé qué aviso de cigarrillos rubios.”
The machine translated version of this fragment is:
“The burning February morning Beatriz Viterbo died, after an imperious agony one moment is not downgraded or sentimentality or fear, I noticed billboards iron of Constitution Square were renovated not know what reporting blond.”
The translation by Norman Thomas di Giovanni in collaboration with Borges, in 1945, is:
“On the burning February morning Beatriz Viterbo died, after braving an agony that never for a single moment gave way to self-pity or fear, I noticed that the sidewalk billboards around Constitution Plaza were advertising some new brand or other of American cigarettes.”
If we compare the machine translated version with the one from the human translator, we can easily see that the trouble begins with the omission of the preposition “on” and is followed by a horribly-confusing attempt at literal translation, such that it is difficult for the reader to understand the original meaning.
The most visible mistake in this case is the confusion resulting from polysemic words and figurative meanings. Another issue that MT can’t overcome is the use of subordinate phrases. Machine translation software today isn’t capable of solving these obstacles by itself, as it lacks the intuition needed to interpret the meaning of the source text.
To provide an equivalent translated version, this fragment requires human interpretation to recreate the mood and the mental image that a mournful narrator may express when facing the fact that life goes on despite death, or when he passes by those “sidewalk billboards around Constitution Plaza” and feels that particular heat of summers in Buenos Aires.
Without a translator proofreading machine translations, the content will be full of mistakes. In many cases, however, the cost of post-edited MT may be higher than just having the content translated by professional translators from the start.