Machine translation (MT) uses a computer algorithm to automatically translate a document from one language to another. The quality of such translations is continuously improving. However, it still cannot fully replace the skills of a human translator.
Machine translation is usually used for specific translation projects e.g. in cases where it is necessary to translate a large amount of data. In every case like this, it is necessary to edit such a translation and proofread it.
The following methods for processing machine translations are available:
Rule-Based Machine Translation (RBMT)
The RBMT translation process is usually based on a dictionary and a set of translation rules. The computer software analyses the source text against the dictionary and the rules and converts the text into the target language. The main benefit of rule-based translation is the predictability of the output. In this way, we have precise control over what is analysed and how it is translated. The preparation of the rule-based system is time-consuming. It requires the time of a linguistic and computer experts to create a working system for a particular type of text for translation. Rule-based machine translation is suitable for expansive translations of a particular type of document that is repeatedly used to create new/several versions and types of products (e.g. expansive documentation for computer software, user documentation, car manuals, etc.).
Corpus-Based Machine Translation (CBMT)
CBMT is based on a bilingual set of documents (source language texts and their translations by translators). Out of this data, the translation system creates a translation table. This table is a form of dictionary containing probabilities (frequencies) for possible individual word or (short) phrase translations. The translation of multiple-word phrases usually produces better results than the translation of individual words. The second part of this form of statistical translation is a language model for the given target language. The purpose of this model is to pick a combination that leads to a meaningful sentence in the translated text.
Corpus-based translation is therefore divided into statistical machine translation (SMT) and example-based machine translation (EBMT).
The main strength of this form of computer-based translation is that it is based on bilingual translation tables. The rules do not have to be defined manually, such as with the rule-based system, and the preparation of the machine translation is therefore less time-consuming.
Hybrid Machine Translation (HMT)
This is a combination of the rule-based and corpus-based translation systems, whereby the aim is to benefit from the strengths of both systems and eliminate their weaknesses. There are many combinations e.g. Lingstat, METIS-II, etc.
Neural Machine Translation (NMT)
NMT is the most recent approach to machine translation. This involves a programmed model that simulates the biological structures of neurons in live organisms and is capable of learning from data via so-called deep learning. In comparison with statistical machine translation, which works with many subcomponents, neural machine translation analyses the sentence as a whole and translates it as a whole. For its successful operation, the neural system needs an extensive parallel corpus. If there is not much data available, the output worsens. The absence of sufficiently large parallel corpora is a problem for many language combinations, in particular for lesser spoken languages (such as Basque or Czech in the worldwide sense), but also even for better known language combinations such as, for example, Russian – German.
Publicly available automatic translators
Examples include Google Translate, Bing Translator, Babylon, Systranet, Prompt, etc. These are not professional machine translation tools. Their main disadvantage is the low quality of the output and the absence of confidentiality with regards to the inserted information (all data inserted into a public automatic translator is automatically stored by the publicly available service of the system). It is for this reason that neither we nor our translators and editors use them – they are contractually bound never to do so.
If you are deciding whether or not to use machine translation technology for the purposes of translating your texts, please do not hesitate to contact us. We can help you decide by telling you for which language combinations and formats machine translation is suitable and how to go about preparing texts for machine translation. If required, we can also help you with such preparations.
Send a non-binding inquiry
Fill in the enquiry form and we will send you an offer within a short space of time
Contact us, send us your text, and we will give you a non-binding price offer
For quick information about our translation and interpreting services, don't hesitate to call us