THE IMPORTANCE OF MACHINE TRANSLATION IN TODAY’S WORLD

WHAT IS MACHINE TRANSLATION?

Machine translation, also called “MT,” is a process performed by a software piece whereby a written text in one language is rendered into another language without humans being involved. It is a sub-field of computational linguistics, a discipline dedicated to comprehend written and spoken language. There are 3 main approaches to MT:

• Rule-based (RBMT): it requires full vocabulary and grammar (in other words, “rules”) of multiple languages to translate autonomously. RBMT is a word-based approach, meaning that it aims at substituting one word in the source language for another in the target language that can work as an equivalent. It was first used by the free online translation service BabelFish in 1997.
• Statistics-based (SMT): it works on a vast amount of textual data; in other words, it examines existing translations made by humans. SMT is a phrase-based approach: instead of translating word-for-word, it works with whole sequences of words (“phrases” or “blocks”). Google launched in 2006 its famous statistical machine translation service Google Translate.
• Neural Machine Translation (NMT): it is the latest stage of machine translation. It consists of neural networks trained to perform translation services. Unlike the other systems, NMT recognizes patterns in the source text and gives a context-based interpretation. It is also referred to as “deep learning” as it analyzes existing translations and learns from them, allowing it to work independently (it mimics how the human brain works). Of the three approaches, neural machine translation offers the highest translation quality, and some claim it reaches a level almost as good as human translation, although these claims are usually overly optimistic since NMT also requires some degree of human intervention.

MACHINE TRANSLATION IN THE LANGUAGE SERVICE INDUSTRY

The translation and localization industries are rapidly growing and, together with them, so does machine translation. In 2019, 13 % of projects for end-clients used MT and, in 2020, this percentage reached 24 %. Content types most commonly translated with MT include technical information, client-internal communications, and third-party content, according to CSA Research.

In relation to market statistics, Grand View Research expects that the global MT market size reaches 983.3 million USD by 2022. Moreover, Global Market Insights expects the global MT market to grow at a compound annual growth rate of 25 % from 2021 to 2027.

However, the increase in the use of machine translation does not mean that translators are no longer required or that they are being replaced. Although MT services are quickly improving, they still rely heavily on human intervention to achieve the required quality for translation. In fact, the process of reviewing and editing machine translation output is called Machine Translation Post-Editing (MTPE).

RECENT ADVANCES

Despite the massive improvements in MT, there are still deficiencies regarding overcoming language barriers. As machine translation systems rely strongly on textual data, low-resource languages (for example, languages that don’t have a standardized writing system) cannot be as accurately translated as others.

Currently, many companies are seeking ways to overcome this hurdle. One of them is the American multinational technology conglomerate Meta. Recently, Meta Artificial Intelligence researchers have announced a long-term effort to build language and MT software with the objective of including most of the world’s languages. To do this, they are building new AI models capable of learning from languages with fewer data. In addition, they are also planning to develop new approaches to speech translation in real time to support languages without a standardized writing system.

RELEVANCE FOR TODAY’S WORLD

Machine translation provides people with numerous advantages. Memsource lists the following:

• Speed and volume: MT services translate thousands of words almost instantaneously.
• Large language selection: MT services offer the possibility of translating to multiple languages (from 50 to 100 languages or more!).
• Reduced costs and faster turnaround: MT can reduce costs and deliver translated projects more quickly (MTPE is cheaper than regular translation and normally it demands less time).
• Automated integration into translation workflows: “translation management systems integrate one or more kinds of MT into their workflow,” extracting content from different sources and turning it into a translatable format for the subsequent application of MT.