Machine translation: what is it and how does it work?

Machine translation is the automated translation of a text into another language using software without the need for human intervention.
Machine translation follows a basic two-step process. The first step involves decoding the meaning of the original text in the source language. The second step encompasses encoding the meaning into the target language. The latest machine translation systems use various artificial intelligence technologies to continuously improve their performance.
The idea of using computers to automatically translate languages first emerged in the early 1950s. At that time, however, machine translation was much more complex than computer scientists had anticipated. It required enormous computing power to process and store the data, which was beyond the capabilities of early machines. It was not until the early 2000s that the software and hardware required became capable of doing basic machine translation. Early developers used statistical databases of languages to ’teach’ computers to translate text. Training these machines involved considerable manual labour and each new language had to be developed from scratch.

Automated vs machine translation: What’s the difference?

Automated translation refers to any automation built into a traditional computer-assisted translation tool (CAT tool). Machine translation is the process of using artificial intelligence (AI) to automatically translate content from one language to another without any human input. Computer-assisted translation is the process of using software to help a human translator transfer the meaning of a written text from one language to another. A CAT tool makes it easier to translate a document between languages by using a number of features such as: translation memory, automatic translation based on glossaries, automatic translation quality checks and so on. With each translation, the translation memory grows, which results in higher consistency and quality, faster delivery of the finished product and reduced costs.

Types of machine translation

There are three types of machine translation: rule-based machine translation, statistical machine translation and neural machine translation.

Rule-based machine translation (RBMT) follows linguistic rules and dictionaries developed by language experts and programmers. It has several serious drawbacks, including the need for a significant amount of human post-editing (manual editing of the machine translation) and generally poor quality.
Statistical machine translation relies on analysing vast amounts of existing human translations to find the closest analogue to the target segment. It builds a statistical model of the relationships between words, phrases and sentences in a text. This system was in use until 2016. The main weakness of statistical machine translation is that it can only translate a phrase if it is present in the reference texts.
The most widely used type of machine translation today is neural machine translation. It uses artificial intelligence to continuously learn new languages and improve the ones it already knows. The software learns and improves with each new experience. However, it is important to note that despite constant improvement, neural machine translation cannot compete with human translation. Proper grammar and declension are often beyond the capabilities of machine translation.