Traditional Translation – Till When?
One day in the not-so-far future, Machine Translation (MT) will be a readily available, fast, and trustworthy service.
Researchers of employment divide all types of occupations into three categories: Creative occupations, caregiving and nursing, and occupations that entail repetition. This last type of occupation will inevitably be replaced by computer programs and robots. Robots or computer software will replace human beings only after learning an entire array of conditions and acquiring the ability to identify the correct path to complete the action, thereby replicating the action that a human would do. Of course, as is the case with navigation software, the advantage of the machine will be its access to multiple updated databases.
It seems that the application of machine translation is closer than ever. True, after witnessing the first attempts back in the 1980s, MT has made great progress due to new technological developments (Big Data, Web Based Apps) and the improvement of models built upon the principles of translation. But we all know that there is still a wide chasm between our status today and machine translations of the future.
But before all translators change their professions, let us first explain the definition of translation and its objective.
What Is Translation
Translation is an act that takes input in one language (the source language) and provides output in a different language (the target language). Every translation follows rules that originate in the world of content of a particular text, its uses and applications. For example, a nautical mile is equivalent to 1800 meters. The conversion of this measurement is a translation. As a second example, the Hebrew "HeChatul [g1] al HaGag" translates into "The cat is on the roof." This translation takes into account a number of parameters that compose the sentence, its meaning, syntax (the position and role of each word in the sentence) and tense. However, this example still cannot be implemented by machine translation because the gender of the cat cannot be identified.
The actions of the translator include gathering background information from the world of content relevant to the source text, and then deciphering the word units (sentences and paragraphs), and transferring them into the target language. These actions include coordination of the complexities of source languages with those of the target language, realizing the differences in the two with regard to syntax, tense, verb structure, accepted measurements and idioms. In effect, the translator must transfer the information into the target language, while maintaining the same meaning as in the original text but in a manner that bridges the cultural differences in the target language.
What Machines Can Do Today
ASR – Automatic Speech Recognition (Speech to Text) – An automatic speech recognition system can convert texts spoken into a microphone into written text on the computer. These engines are already prevalent in mobile phones to dictate text or issue commands. In the past, these engines were based upon a comparison between a pre-recorded text and text spoken into the microphone. Today, ASR engines separate syllables (phonetics) and even in cases without 100% word compatibility, they can review a number of words and then use the process of elimination to choose one that fits the preceding context. ASR can also correct spelling based on context or predict the next words in a sequence by following basis algorithms.
MT – Machine Translation such as Google Translate depends upon the amount of information made accessible (texts that were already translated and adapted to various languages) and enhanced by users over time. The advantage of MT is its familiarity with a great number of words and their translations, models of sentence structures, measurement conversion and so forth. Longer source texts will be translated more precisely by machine translation, as information from within the text improves the translation of information from external sources.
AI – Artificial Intelligence – This broad sector utilizes models and algorithms to identify one correct answer or make a decision similar to one a human would make. Computerized translation includes the subfield of NLP which has developed over the past few years in a computerized environment, thus providing the ability to deduce information from language and define its significance. NLP is a scientific field that has been in existence for decades but the transfer of the medium to the web makes it accessible and enables it to serve and enhance translation. Use of AI helps translation software understand what is being said.
Translation and Artificial Intelligence share other joint components, many still in the conceptual research and development stage, for example – the recognition and removal of cynical undertones in language during the comprehension process of the text.