Document home

Machine Translator

Machine Translation

Machine translation (Wikipedia) is a technology where a cloud service or local application automatically translates sentences from one language into others. Machine translation lets you quickly translate your strings for testing purposes or for post-editing. The quality of machine translators is getting better year by year.

Supported machine translation services

There is no single machine translation engine that is good in all languages. Each engine has good language pairs and not such good pairs. This is why Soluling has interfaces to several machine translation services. They are:

Amazon Translate

Amazon Translate

Amazon Translate (homepage) is a neural machine translator service provided by Amazon. Amazon Translate is a paid service. However, if you are a new AWS customer, you can use it for free up to 2 million characters per month for the first 12 months. If you use more, you have to pay. After 12 months you have to pay a flat rate per million characters. To use Amazon Translate, you have to configure it.

DeepL

DeepL Translator

DeepL Translator (Wikipedia) is a neural machine translator service. DeepL has a good translation quality, but it supports fewer languages than other machine translators. DeepL is a paid service. You pay a flat monthly rate that includes unlimited use of DeepL translation. To use DeepL, you have to configure it.

Google Translate

Google Cloud Translation

Google Cloud Translation (Wikipedia) is a neural machine translation service provided by Google. Google Cloud Translation provides a good translation quality, and it supports more languages than Microsoft Translator. Google Cloud Translation is a paid service. It cost $20 per million characters. You don't have to choose a flat rate subscription, but you are charged for the actual usage. To use Google Cloud Translation, you have to configure it.

IBM Watson Language Translate

IBM Watson Language Translator

IBM Watson Language Translator (homepage) is a neural machine translator service provided by IBM. The translator is a free service of up to 1 million characters per month. If you use more, you have to pay. To use IBM Watson Language Translator, you have to configure it.

Microsoft Translator

Microsoft Translator

Microsoft Translator (Wikipedia) is a neural machine translator service provided by Microsoft. It provides dozens of languages. Microsoft Translator is a free service up to 2 million characters per month. If you use more, you have to pay. In order to use Microsoft Translator, you have to configure it.

ModernMT

ModernMT

ModernMT (homepage) is a neural machine translator service. ModernMT has a good translation quality. ModernMT is a paid service. You pay a flat rate for every million characters. You will need either Human-in-the-loop or Professional ModernMT subscription. To use ModernMT, you have to configure it.

Google Translate

MyMemory

MyMemory (home page) is a translation memory and machine translator. MyMemory is a free service of up to 10 000 words per day. The quality is good if there is a close match to an existing translation memory item. If not. the quality is not so good.

Yandex Translate

Yandex Translate

Yandex Translate (Wikipedia) is a machine translator service provided by Yandex. It provides dozens of languages. Yandex Translate is a free service up to 10 million characters per month. If you use more, you have to pay. Yandex Translate uses hybrid neural-statistical machine translation for some language pairs and statistical for most language pairs. To use Yandex Translate, you have to configure it.

When you install Soluling, all the above machine translator engines are added, but most of them are not configured. To configure them, you have to get your account key and enter it to the key field of the engine. Use the Machine Translator sheet on the Options ribbon to configure the common machine translation engines. Each project can then either use the configured common machine translators or configure its own project-specific translators or use both. Use the Machine Translation sheet in the project dialog to edit machine translation settings of a project.

Interactive vs. Batch mode

You can use machine translation in two ways. The first one is an interactive mode where whenever you move to a new string, Soluling automatically gets the translations using the configured machine translator engines and shows the translations in a candidate window where you can select the translation you want to use. The second more is the batch mode, where you use Soluling's menu to start translation untranslated strings using the selected machine translators.

Models

Several types of machine translation technologies exist. Phrase-based approaches (PBMT) and traditional statistical machine translation (SMT) has been mainstream for several years. However, recently, Neural Machine Translation (NMT) has become the mainstream because it provides better quality. Most machine translation service providers have switched from traditional machine translation to neural machine translation. Some services use a hybrid model where both neural and statistical models are used at the same time. Some engines let you train your neural model and use it instead of the standard neural model.

Quality-Translate

Machine translation quality is sometimes not very accurate. The results are getting better but do not meet the quality of the human translations. To increase the quality of machine translation, Soluling can translate the sentence from multiple sources (e.g., using Google and Microsoft). Soluling will then analyze those results and selects the best translation. The translation status and match percentage depend on how well the different translations match. If all sources provide similar results, then the translation status and match percentage are higher. This technology helps you to filter out bad quality machine translations. The technology can be used if you have added at least two machine translators in the Machine Translators sheet.

Use Quality-Translate to choose Use Quality-Translate menu instead of a machine translator specific menu (e.g., Use Microsoft Translator).

Removing machine translations or post-editing them

Because machine translation quality is less than optimal, it may not be used as final translations. Most often, machine translation is used during the development and testing phase, where you need to have real translations quickly. Later on the process, machine translations will be edited or replaced with real human-based translations. To make this process easier, Soluling can remove machine translations. To remove machine translations right-click the column and choose Machine Translate | Remove Machine Translations.

If you edit a translation currently having a machine translation, the translation will become real and won't be removed when removing machine translations. Editing means adding or removing characters or changing the translation status.

Project specific settings

When using the interactive machine translation, Soluling uses only those machine translators that are currently enabled for the project. You can configure the settings from the project settings. Right-click the project root in the project tree and choose Options. Then select the Machine translation sheet. The following image shows a situation where you have several machine translators configured in Soluling, but only Microsoft is enabled for this project.

Project's translation memory settings

The lower half of the sheet lets you configure machine translators that are specific for this project only.

Soluling uses machine translation when performing back-translation.

 

What is the difference between a termbase, translation memory, existing translations, and machine translator?