Q3 Product Release Notes
As the world continues to evolve, so do customers’ needs. With that in mind, Unbabel sharpened its focus on enhancing customer experience over the course of Q3. We’ve expanded our machine translation capabilities and made improvements across CRM integrations, the Unbabel Editor, and much more.
World Renowned Quality Estimation
The Unbabel R&D team recently took part in the World Machine Translation 2022 Translation Quality Estimation shared task: Of the 19 institutions that participated, Unbabel’s Quality Estimation technology ranked first in all categories. When evaluated based on its ability to predict the quality of various neural machine translation outputs, the performance of Unbabel’s Quality Estimation outranked all competitors, further validating that our technology is the best in the industry.
Our customers can rest assured knowing that Unbabel, as their provider, is the global leader in Quality Estimation. A detailed findings paper analyzing the results of this shared task will be published later this year at the EMNLP 2022 conference.
Integrations
The Unbabel team released two features to enhance our multilingual CRM Integrations solutions.
-
Zendesk Messaging: Enhances the Agent Workspace experience by enabling support through social channels like WhatsApp and Messenger. Learn more about the benefits of Zendesk Messaging.
-
Salesforce Service Cloud UI and UX update: Streamlines the agent view so the entire case appears in the agent’s native language. Agents can now seamlessly interact with customers as if all parties were speaking the same language and have the option to expand to view the complete conversation in both languages.
Unbabel for Zendesk Messaging, now enabling support on social channels, and the consolidated Salesforce feed are great examples of how we plan to continuously expand our multilingual customer experience offering. And we’re just getting started.
Machine Translation
Unbabel recently released a new function of machine translation (MT) aimed at enhancing our current technology with marketing content translation capabilities. By building a new series of domain adaptive models to adjust the way we train bilingual MT models, customers can experience a high standard of quality for more than 100 language pairs in the marketing content use case.
In Q3, we also released Large Pre-trained Models, which enable us to serve multiple language pairs from one single MT model and scale to new languages faster. Unbabel customers benefit by having one machine translation model translate a single source language into different target languages, and now have access to many additional language pairs.
Unbabel Editors
Additionally, the Unbabel team implemented processes and features to enhance the capabilities of Unbabel editors. One of the more noteworthy changes involves a more thorough screening process for editors and translators, to classify them based on their abilities and to match them with the translation project they’re most qualified for. The in-depth evaluation gathers information to ensure translations are routed to the right editor, and in turn, provides an additional layer of quality assurance to our customers.
We also released a number of features aimed at enhancing the Unbabel editor’s workflow, including Contextual Notes, which helps customers add context and instructions at the segment or sentence level. This feature opens up a world of possibilities in providing context-related messaging from the customer to the translator.
Tag Handling was also released this quarter which focused on redefining our approach to file formatting. Editors now have the autonomy to adjust file formatting in order to ensure our customer receives a perfect finished product.
This past quarter, Unbablers were on a roll — delivering features and enhancements with customer experience top of mind.
We will continue taking steps to ensure that we’re providing our customers with the best solution in the market.
If you have any questions about what we’ve been working on, and how it can help your business improve its multilingual CX, contact us here:
