What is Machine Translation?

Machine translation plays an important part in language translation around the world, and is used by thousands of individuals and businesses every day. Machine translation, often abbreviated to MT, is a sub-field of computational linguistics where computer software is used to translate either text or speech from one natural world language to another. While machine translation is still a long way from being perfect, increasing research and opportunity means that it is more accurate than ever before.

At its most basic level, machine translation performs a simple one to one substitution between words and recognised phrases. However, some more advanced translation software is able to perform more complex procedures and customise translations based on a number of contextual relationships.

The Challenges of Machine Translation

While the idea of machine translation has been around for a long time, it has only been in use practically since the late 1980s. This is due mostly to advances in computer processor speed during this time, and growing interest in statistical models for the development of translation software solutions. The machine translation process involves two distinctive steps: decoding the original meaning of the source material, and re-encoding this meaning into a target language. However, while this process may seem simple, there is much complexity behind the analysis and cognitive procedures that are necessary for accurate translation. Some of the knowledge that must be recognised and addressed by the machine translation process include grammar, semantics, language idioms, syntax, and cultural meaning.

When a machine is involved in the language translation process, there are a number of challenges that have to be faced, especially with regard to language structure, cultural norms, and specific idioms. There are a number of different methods that are used in machine translation, most of which can be categorised as rule based, statistical based, or example based paradigms. Most machine translation uses a hybrid approach that utilises many of these methodologies, including transfer-based, interlingual, and dictionary-based translation rules. In hybrid systems, rules can either be post processed by statistics or statistics can be guided by rules. While machine translation systems still need to face a number of challenges, they are sure to play a major role in the future of language translation around the world.