Q1 & Q2 2020

 

In the first half of 2020, we enhanced the API methods and a couple of features in the Translation Editor, added support for new file formats and languages, and focused on the MT integration.

 

  • We have developed a new API 2 method to calculate statistics on segment and text fields. See API2 Aggregations for more details.

  • In order to cost Flex or crowdsourcing work, we use a set of API methods to calculate counts and costs and create invoices. We have recently added API2 methods to calculate word counts and costs by period and worker.

  • There is a new option available in the Flex filter configuration that tracks content modifications: “Do not reset the status of translation with markup-only changes". This means that the status of the translation is reset only when there are actual changes in the target text. The Flex filter helps translators quickly discover the changes that have been made to the target text.

This filter applies only to the target text! Any modifications in the source text will trigger the red status in the translation.

  • We have enhanced the response time of API calls for job lists. The API now excludes jobs with the status "Archived, Cancelled, Completed" if they have been created more than a year ago.

  • The Flex log file now includes additional information:

    • Commits are word counted and counts are included in the log.

    • Invalid segments that reference nonexistent segments in the target text are logged with the bad keys.

    • Event is included in the log file even if the commit did not affect any segment in the Flex document.

    • The unique transaction ID is logged with each commit.

  • A new API method has been added to append custom messages to the Flex log. In certain integration scenarios, you may want to add additional log entries. See API2 > Events logs for more details.

  • Double-clicking a .wbflex.log file in the project documents page now opens a web page that shows commit details in a compact and easy-to-understand format.

  • We have enhanced the XLIFF file format configuration to support the CDATA (character data) elements. If your source file contains CDATA elements, the target files are now generated correctly.

  • We have added Google Purple (#9900ff) to the list of colors available in the configuration of the MS Word file type. Go to Customization > Microsoft Word - Configuration > Do not translate ticked colors and select Google purple from the list.

  • Wordbee can now handle the European Broadcasting Union (EBU) STL subtitle format. This file format is widely adopted in the broadcast industry.

  • JSON2 filter is now the default filter when creating a new JSON configuration. This change does not affect existing configurations.

Wordbee Translators supports a total of 635 languages around the world. We have recently added the following languages:

  • Javanese

  • Maay Maay (ymm)

  • Mandinka (mnk)

  • Runyankore/Nkore

  • Sundanese

  • Sindhi

  • Tetun (tdt)

  • Javanese

  • Maay Maay (ymm)

  • Mandinka (mnk)

  • Runyankore/Nkore

  • Sundanese

  • Sindhi

  • Tetun (tdt)

The system shows which MT engine was used for translation

Wordbee Translator integrates via API with different Machine Translation systems. With the help of the Meta connector available in the MT Customization area, you can select your favorite MT system for each of your language pairs. The system now stores this information in the project properties. For example, when you hover over the MT icon of a machine-translated text in the Editor, you will now be able to see which MT system was used to produce the raw translation. This information is recorded in other parts of the project as well:

  • the revision history of MT translated texts;

  • the word count popup of a document/job;

  • the word count detailed report;

  • the Codyt job works progress tooltip.


Tilde Neural Machine Translation is now accessible from Wordbee Translator

Wordbee Translator currently supports ten machine translation systems. We have recently added Tilde Neural Machine Translation (NMT) which is based on neural networks. Research has shown that neural networks-based systems produce more fluent translations than other machine translation architectures. According to Tilde, their NMT system takes into account word reordering, dependencies, morpho-syntactic agreements, and other features that are not handled proficiently by statistical machine translation.

To enable the Tilde NMT connector in the MT Customization area of your platform, you first need to obtain an API URL and client ID authorization token from Tilde. Finally yet importantly, as of June 18, you can configure the API URL manually as well. This recent development enables you to switch easily between Tilde’s infrastructure (for example, to compare different tools) and your own infrastructure. In conclusion, it is possible to have more than one Tilde MT systems configured with different URLs.

 

 

 

 

  • It is now possible to configure a new QA rule to flag missing spaces before preconfigured punctuation marks. This new QA rule is deactivated by default. We recommend that you enable and properly configured it only for those languages where mandatory spaces before punctuation are required such as French, for example.

Learn more

Have you not found what you were looking for? Check our Release History and Knowledge Base.

 

Copyright Wordbee - Buzzin' Outside the Box since 2008