- What is Machine Translation?
- What types of MT are there?
- When can it be used?
- How can it help reduce my company costs?
1. What is Machine Translation?
Machine translation (MT), not to be confused with computer-aided translation (CAT), is a sub-field of computational linguistics that investigates the use of software to translate text or speech from one language to another, through the use of AI (Artificial intelligence).
Dr. Jason Brownlee provides the following description on his website: “Given a sequence of text in a source language, there is no one single best translation of that text to another language. This is because of the natural ambiguity and flexibility of human language. This makes the challenge of automatic machine translation difficult, perhaps one of the most difficult in artificial intelligence:
The fact is that accurate translation requires background knowledge in order to resolve ambiguity and establish the content of the sentence. (Page 21 of the book: Artificial Intelligence, A Modern Approach, by Stuart Russell & Peter Norvig)
According to the Research Group on Statistical Machine Translation at the University of Edinburgh: “The dream of automatically translating documents from foreign languages into English (or between any two languages) is one of the oldest pursuits of artificial intelligence research. Armed with vast amounts of example translations and powerful computers, we can witness significant progress toward achieving that dream.”
2. What types of MT are there?
According to Dr Jason Brownlee, “Automatic or machine translation is perhaps one of the most challenging artificial intelligence tasks given the fluidity of human language. Classically, rule-based systems were used for this task, which were replaced in the 1990s with statistical methods. More recently, deep neural network models achieve state-of-the-art results in a field that is aptly named neural machine translation.
Statistical Machine Translation:
According to the Statistical Machine Translation organisation, statistical machine translation “is the translation of a text from one language into another, by a computer that has learned to translate on the basis of many translated texts,” through the use of statistics. For example, in the diagram below, it can be seen how this type of MT works (example taken from a tutorial on Statistical Machine Translation by Kevin Knight and Philipp Koehn, from the University of Southern California):
It is, therefore, a type of Machine Translation that is added to and improved through incorporating new translations.
Neural Machine Translation
According to Wikipedia, is an approach to machine translation that uses an enormous artificial neural network. It differs from statistical translations based on sentences that use separately designed sub-components. Translation services like Google, Yandex and Microsoft currently use NMT. Google uses Google Neural Machine Translation (GNMT) in preference to statistical methods used in the past. Microsoft uses similar technology for its voice translators (including Microsoft Translator live and Skype Translator). Harvard NLP has launched a system based on neural machine translation open code – OpenNMT (https://nlp.seas.harvard.edu/) .”
Neural machine translation is having very positive outcomes and is currently being used most often.
We also use a machine translation service at LocalizationLab, known as Softcatalà. It’s a non-profit association that aims to encourage the use of Catalan in all areas of new technologies (TIC). https://www.softcatala.org/traductor/
3.When can it be used?
Machine Translation doesn’t always work properly, and the project needs to be analysed thoroughly to decide whether it will be useful or not. At LocalizationLab, we use neural machine translation tools when we are sure that it can reduce cost and time without affecting overall translation quality. This article talks about one of our success stories, when we used machine translation for technical and legal texts.
Before suggesting that a customer uses this service, first we analyse the topic at hand, the project size and the language combinations. We also do a test with an extract of text that is then revised by one of our translators to decide if the outcome is optimal.
Importantly, there is always a revision stage after using machine translation, carried out by a native, specialised translator. This is known as Machine Translation Post-Editing (MTPE). It consists of a comprehensive revision and shouldn’t be confused with the type of revision applied to a text translated by a professional translator. MTPE requires a very detailed revision and expert knowledge of the text’s topic.
4. How can it help reduce my company costs?
Machine translation can help reduce translation costs, and it can be used to translate a high volume of documentation, in several languages, at a lower cost. But it needs to be done properly. Not all translators want to work on post-editing. If the machine translation is of poor quality, the person revising the document can waste a lot of time.
The role of the person revising the post-editing is incredibly important in the process, as they will be the only person who revises the text. They need to ensure that it is 100% accurate. Detecting errors in a machine-translated text is difficult. In fact, it’s easier to spot a glaring mistake than realise there is a missing “not,” which could have serious consequences in a contractual clause.
If you have extensive content that you need translated into other languages, or very long and repetitive documents, machine translation can probably help do what you require more quickly and economically. Why not drop us a line (firstname.lastname@example.org) and we’ll estimate the cost and time so that you can take it into account?