Human Translation Vs. Machine Translation: A Triumph of Nuance over Nonsense?

Posted on 06/18/13 in INDUSTRY NEWS, LANGUAGE, TRANSLATION, No Comments

Latin and Greek became “universal” languages for a time in Western civilizations.  However, their dominance and popularity waned as the Age of Discovery fostered the prevalence of the most widely spoken European languages in the newly acquired colonies around the world.  Despite having been in existence for well over a century, the artificially conceived language, Esperanto, has not yet become a secondary official language of any world nation, nor is it in significant international usage as a lingua franca, despite being far easier and faster to learn than English, which continues to expand its dominance as a global first or second language.

Seamless inter-species communication is ensured in popular science fiction such as Star Trek, for example, through the presence of universal translation devices either as implants or built into communication devices possessed by both aliens and humans alike, neatly sidestepping the issue of how a tentacled creature from the farthest regions of the galaxy speaks perfect English.

Here on 21st century Earth, however, machine translation is still in its infancy.  Real research into the provision of automated translation began in the 1940’s and the advent of the internet accelerated the development of software for this purpose such as SYSTRAN and Babelfish.  However, scepticism remains strong amongst humanoid translators (!) as to whether non-sentient software can effectively and indeed accurately reproduce a piece of text into another language.  It is not sufficient to merely exchange words in one language for equivalents in another: a translator’s knowledge is not simply derived from having learned the native and target languages to a high level of proficiency.  A good translator will have lived in the countries where the respective languages are spoken, will continue to read extensively, will be aware of cultural and linguistic nuances and will keep up to date through various media with current affairs, the introduction of new words and phrases, and most importantly will have a feeling, an intuitive sense for the languages.  No two translators will arrive at precisely the same result, particularly where a more complex sentence or piece of writing is involved.  Just as all language speakers formulate speech differently, all translators will likewise draw on their own personal knowledge in addition to their linguistic competence to render the text in a unique way.

To demonstrate, DCU Language Services asked some of its translators to provide their own rendition of a German sentence in English.  The same sentences were also translated using some of the best-known translation software available on the internet: Google Translate, SYSTRAN, Bing (Microsoft Translator) and Babylon. From the outcomes below, it is possible to see the differences between the various translators’ interpretations of the original text, but also to observe the shortcomings of the translation software:

Example 1:

Original sentence: Dem Leasinggeber steht es frei, zur Wahrnehmung dieser Aufgaben und seiner Interessen seinerseits einen sachverständigen Dritten zu beauftragen.
Translator 1: The lessor is at liberty to retain the services of a competent third party for the performance of these tasks and protection of its interests.
Translator 2: For his part, the lessor is free to appoint a knowledgeable third party to execute these duties and represent his interests.
Translator 3: The lessor shall be at liberty to entrust their interests and decisions on their part to a competent third party.
Machine translation 1: The lessor is free to contract for the performance of these tasks and its interests, in turn, an expert third party.
Machine translation 2: It is open the leasing giver to assign for the perception of these tasks and its interests for his part an expert third.

Example 2:

Original Sentence: Der Widerspruch ist schriftlich oder mündlich zur Niederschrift binnen eines Monats nach Bekanntgabe dieses Bescheides bei der im Briefkopf genannten Stelle einzureichen.
Translator 1: The appeal must be submitted in writing or orally for recording in writing within one month following issue of this notification at the offices specified in the letterhead.
Translator 2:  The appeal is to be submitted either in writing or orally to be put on record within one month after publication of this Notice, to the office designated in the letterhead.
Translator 3: Objections are to be filed within a month of disclosure of this decision to the person named in the letterhead, verbally or in writing.
Machine Translation 1:  The contradiction is in writing or orally to the transcript within one month of the publication of this notification in the specified place in the letterhead submitted.
Machine Translation 2: The contradiction is to be submitted in writing or verbally to the minute within one month after publication of this answer with the place specified in the letterhead.

While it cannot be denied that mechanical translation has improved in recent years, the samples above show that there is still too much scope for erroneous and, in some cases, risible results, and that consistency is unreliable, particularly when larger texts require translating and it is important for the target audience or reader that a specific style or nuance is required.

In short, there are many aspects of modern life in which automated technology can improve simple tasks and provide an outcome of superior quality, however, until it is possible for a computer to write a book, a stanza of poetry, a personalised letter or a slogan for a marketing campaign, it is best to leave translation to those who do it best.

Thanks to our translator Caroline Handschuh for contributing this article to our blog, and to our other translators for assisting with the translations: Oli Gallagher, Treasa Uí Mhuircheartaigh and Sarah Harrison.