How to machine translate with Intento API
Wordbee Translator now integrates with Intento API, which offers one-stop access to a wide range of third-party machine translation systems. Customers can connect to 29 MT providers and models through a single interface, and evaluate the engines for specific projects.
How to get your Intento credentials
To be able to use the Intento API in Wordbee Translator, you will need to sign-up and get a Production API Key from Intento.
Intento Web Tools
Besides statistics, you will also be able to use the Intento Web Tools to translate files and test the different MT models directly in your Intento Machine Translation Web interface. The service will display the MT models based on your language pairs.
What is smart routing?
You can use the smart routing feature to route your translation request to the best MT provider. Intento selects the best MT model for your text and language pair based on the following information:
apriori benchmark on the standard dataset
provider usage statistics
To learn more about this feature, see the Intento Smart Routing article.
Configure your Intento connector in Wordbee
When you have the Production API Key, go to Wordbee Translator > Settings > Machine Translation Systems and configure the Intento API. See Machine Translation Settings for more details.
Select Intento and open the configuration options.
Fill in the name.
Enable the full feature set.
Select the connection parameters:
Leave the Intento URL blank because this will be updated automatically when you fill in your Key.
Fill in the Intento Key you received from the provider, and click on Check key and reset button to validate the credentials.
You may set up different Intento profiles with the same API Key, for example, Intento DeepL, Intento Baidu, Intento Google, etc. Hence, when you mark documents for translation in Wordbee Translator, you will be able to select different Intento profiles for each language pair in your project.
In the Provider section, you can choose between two options:
Option 1. Select a Specific MT provider from the list. For demo purposes, we have selected DeepL API v1.
Check if the MT engine offers the language pairs you want to use in your project.
Option 2. Enable the smart routing feature, which automatically selects the best engine for your language pair.
Note that the smart route option does not provide information about the supported language pairs. You will need to add a new Meta Connector for a previously created Intento profile and add the languages of your project. Then, go to your Word Count Profile that you have specifically created for the MT workflows, and choose the Meta Connector as a proxy to the underlying Intento connectors (for all the different languages).
How to use Intento in your translation projects
In Wordbee Translator you can create two types of projects:
Standard projects. Use this type of projects when you translate a series of documents within a specific timeframe.
CoDyt projects. Use this type of projects when you need to set up an automated continuous translation workflow.
Access the embedded links to learn how to create each type of project.
Set up a Post-editing workflow with Intento in CoDyt projects
To set up a Post-editing workflow with Intento in a CoDyt project, proceed as follows:
Configure and enable the Intento profile you want to use.
Make sure you have a Post-editing task available in Settings > Task categories. Otherwise, you need to create one.
Create a Word Count profile specifically for an MT PE workflow and enable the Pre-translation by machine option.
Create a CoDyt project and attach the Word Count profile you have just created.
Select your workflow steps in Workflow & Suppliers, for example, Post-editing, Review.
Attach domain-specific translation memories and terminology databases.
Upload your documents and mark them for online translation. The configuration window opens.
Enable the Machine translate text not otherwise pre-translated.
Click on Configure to select the Intento profiles you have enabled.
To open the segments in the Translation Editor, right-click on the document and select View segments.
For step-by-step instructions, please see this article:How to use TM and MT to translate documents in CoDyt projects
The example below shows that the majority of segments (green colour) have been pre-translated with the TM attached to the project. The orange target segments come from the Intento MT engine. The translator needs to check the quality of the raw output and post-edit the segments if needed. When the editing is finished, press CTRL+ENTER to confirm the translation.
If you would like to try another MT engine for a specific segment, use the MT tab on the Translation Finder widget to select the provider from the drop-down list. Remember that the MT system needs to be enabled first!
Set up a Post-editing workflow with MT Hive
Wordbee MT Hive enables customers to request an instant translation via the Client Portal. Workflows can be automatically kicked off in order to reduce management tasks and help the customer save time and costs. Depending on the quality of the MT output, the customer may download the files or request a post-editing job. See How to manage MT Hive requests to learn how customers can request MT post-editing and how you can set up the workflow.
Use Intento to translate full documents
To use Intento to translate full documents within a project, see this article: How to use MT to translate full documents in a project. The example below shows which text you can choose to pre-translate with the selected Machine Translation:
Untranslated segments only. Untranslated or machine translated segments. All segments. If this option is selected, the MT system will translate only those segments that either contain no translation or they have been translated by another MT engine.
Never translate segments in green status. If this option is selected, the MT system will not translate the segments that you have already translated and confirmed.
Never translate locked segments. If this option is selected, the MT system will not translate the locked segments.