Microsoft wants to help organizations hit the ground running as they attempt to build custom language translation services, instead of starting from scratch.
Available now via the Microsoft Translator Hub are three new customization options for users of the Translator application programming interface (API), the company announced today. “Today, we are changing how companies approach automatic translation by leveraging our artificial intelligence expertise to enable anyone to quickly and easily customize translation systems, even without large amounts of previously translated sentences. In addition, we are giving you the chance to progressively improve the system as more data becomes available,” said Microsoft’s Translator team in a Jan. 27 announcement.
Microsoft Translator is a cloud-based language translation platform that allows organizations to add text translation services to their applications. Last March, the software giant announced that Translator had crossed a major milestone by supporting 50 languages.
Today, Microsoft is offering an expanded set of Translator capabilities, enabling organizations to fine-tune their own translation systems. Included in this round of updates is the ability to select from two new categories, Tech and Speech, to help users customize the context of their translations beyond Translator’s default “by narrowing the scope of the statistical analysis that Microsoft Translator uses to translate your text,” stated the company.
Users that go with the Speech option will have access to some of the technology that Microsoft’s uses in its own Skype Translator real-time translation software.
“The ‘speech’ category was developed in the last 18 months as we built Skype Translator,” said Microsoft’s staffers. “For Skype Translator to work properly, it was critical to be able to translate spoken text, which in most cases can be very different from the written text.”
Currently, it supports the same seven spoken languages as Skype Translator and the mobile Microsoft Translator apps for Android and iOS, namely Chinese Mandarin, English, French, German, Italian, Portuguese and Spanish. “As new speech languages are released for these applications, the equivalent ‘speech’ category will become available for text translation in our core Translator API as well,” pledged the Translator team.
Translator also now allows customers to upload custom dictionaries, using word lists in simple Excel spreadsheets to train the system. “Dictionaries allow you to make your own foreign language word lists so that the terminology that is unique to your business or industry will translate just the way you want,” stated the group.
Finally, Microsoft has added more flexibility in the number of parallel sentences (which Microsoft refers to as “pre-translated sentences in the original and target language”) required to train the system. “As was possible since the Hub launched, but now starting with only 5,000 sentences rather than 10,000 previously, you can use any amount of parallel sentences above 5,000 to customize your translations. With more than 5,000 parallel sentence, you can begin to create a system that is learning new terms and phrases in the right context and tone of your business,” stated the company. Users can start with at least 1,000 parallel sentences, though the quality of the translations may not match those with higher sentence counts.