Pangeanic's ECO: a mission to translate and anonymize all the text in the world

The importance of translation is reflected in numerous day-to-day applications, from mere entertainment to e-commerce services and much more sensitive and complex applications of legal, medical, political or industrial importance. A good example of this is the recent translation of the popular series Squid Game, Netflix's most watched show, which has been criticized for poor use of technology and human review in its subtitles.

Various media sources, including an article from Fotogramas on its huge translation errors, show how the meaning of the original script has been almost completely lost in the translation of the subtitles into English. The key lies in the close link between people and the reliability of the data produced by the translation machines, i.e., human expectations and skills when it comes to using the technology efficiently. There is also a link between visible use and free systems.

The new hybrid forms of translation make use of the most pioneering technologies, such as machine learning and deep learning, to deliver a new generation of adaptive systems that close the gap with human parity.

ECO: a new form of deep adaptive machine translation

The facts show that what is not scalable is not sustainable, and traditional methods of human translation or using CAT tools (Computer Aided Translation, translation memory-based systems) are incapable of addressing scalability in the human quality of translations. As a result of this problem, the private cloud document translation platform, ECO, was created, and its subsequent v2 manages to combine human intelligence with state-of-the-art machine translation technology.

The platform offers a new form of adaptive machine translation featuring customized cognitive engines. Based on artificial intelligence (AI) systems, it is able to mimic the terminology and style of users in a matter of minutes or even in real time, creating translation engines whose results offer a quality close to human parity.

The platform implements the highest-quality translations and anonymization and security techniques for businesses, either on-premise or from Pangeanic's private cloud of servers. Here's how we intend to accomplish our mission, combining artificial and human intelligence to translate or anonymize all types of content with ECO.

How to translate all types of content with the ECO platform

Pangeanic's ECO platform can translate documents, Excel files, PowerPoint files, terms and conditions, product catalogs, emails, conversations, subtitles and any other content imaginable. In addition, the anonymization services help organizations to share information in the most secure way and to comply with privacy laws.

Whether you are looking for a fast translation service, a stand-alone anonymization system, or a recurring translation and anonymization service using SaaS (Software as a service), ECO’s Natural Language Processing (NLP) ecosystem allows customers to choose a single language process or integrate them all. ECO is available in monthly subscription packages available according to the needs of small, medium or corporate companies.

 Machine translation through an API is instantaneous and ECO automates workflows so that neither users nor Pangeanic have to spend time on management. All subsequent steps are self-selectable by the user and can even be programmed for a more efficient transfer of content.

• Request: the request for machine translation of the document or text fragment, with or without anonymization. The user uploads the documents from his local machine to the platform and selects the options. Advanced or corporate users can choose to send programmed batches, with self-identification and encryption between systems for constant information delivery. 
• Similarity analysis (optional): the first step is to compare whether there are sentences previously translated and reviewed by humans to maintain the style, terminology and save costs for the user. The client's translation memory is checked and matches and similarities between previous translations are extracted.
• Adaptive machine translation or generic machine translation/anonymization: ECO chooses the best specialized translation engine by analyzing the type of incoming content, or the client's translation engine that has been previously customized for the client. In the case of anonymization, the language and specialized machine are chosen where appropriate. The anonymization process can be carried out prior to translation to meet privacy requirements or on its own (without translation).
• Human editing of content (optional): if the content to be translated is intended for publication, that is, it is aimed at an audience that will read it, we recommend a review by our human editors, who are selected based on previous qualifications, experience in the subject matter and availability. We call this step “human-in-the-loop” and it is a completely optional step. When content is not essential (reviews, comments or opinions on the web, purely informative documents, patent information, understanding news, etc.), it is neither optimal nor scalable. If a human review is carried out, it is stored in the client's translation memory and would be applied in Step 2 in future jobs.
• Evaluation of quality and final result: ECO produces quality metrics collected at each stage of the translation process, including both general rules and those specific to each client's requirements. The combination of technologies allows the final result to be a translation of almost perfect quality (close to human parity). These evaluations are part of the quality and consumption metrics and are available to customers in their profile.
• Automatic sending (API or download): documents are sent according to the customer's requirement, automatically via the translation API or by private download from the user's profile.
  
  

ECO self-learning network

The magic of the ECO process is that the more you translate, the better the results of the system are. As the platform learns the preferences of its customers, translation memories are expanded, and translation engines are improved to bring them closer and closer to human parity. If the client does not use this step, they can always provide data or point to data sources for Pangeanic to perform this mining work and specialize the engine.

Thanks to machine learning, which allows for more and more contextual accuracy and specialization as it becomes more familiar with the texts of a given source, organizations and companies that outsource translation and anonymization as SaaS find in Pangeanic a technology partner that offers highly sophisticated and scalable translation resources.

ECO: the balance between human intelligence and Natural Language Processing applied to translation

The demand for professional translations is growing exponentially. The communication systems of large companies need to know the opinions of their users, which are expressed through conversations in multiple languages. The original sources of information are multiple and do not necessarily occur in a familiar language. The fragmented system of handcrafted production offered by the thousands of translation agencies around the world is now obsolete. Pangeanic identified this need and changed its mission to become a Natural Language Processing technology company focusing on the generation of state-of-the-art machine translation and support systems such as anonymization. Our mission and vision were supported by several projects carried out for the European Commission (e.g., MAPA-project.eu for anonymization, NEC-TM for translation memories, NTEU for translation engines). Each project has been a step in our growth and development and has finally materialized in the satisfaction of customers and users around the world.

As Pangeanic’s CEO explains to PR Newswire, ECO's AI processes help increase productivity, gain insights from data, reduce costs and simplify enterprise content management; all of this always complying with the high standards of security and privacy.

In this way, ECO fulfills our mission in combining artificial and human intelligence to obtain actionable insights thanks to its near-human neural machine translation solutions, with deep learning and translation engines that understand the surrounding context in tone, style and terminology, and detect actors and entities as needed to automate anonymization and comply with privacy regulations in different jurisdictions.

We believe in process automation and that, in many scenarios, Pangeanic's machine translation technology performs a job at a price and speed that a traditional translation service cannot. In some cases, the contribution of the "human touch" is still necessary, with the difference that our approach is aimed at correcting and editing the work of AI which improves with use rather than taking on the translation as a whole. We teach machines human language by collecting millions of parallel sentences across languages. In this way, we favor scalable processes, both in quality and in accuracy and privacy. 

ECO by Pangeanic: hybrid translation engines with near-human quality

ECO has become the document translator with anonymization solutions par excellence, based on deep learning and machine translation. It offers translation engines for local implementation or complete machine translation systems according to the needs of each client.

With our translation and anonymization SaaS tool, we offer solutions to clients in sectors as diverse as pharmaceuticals, healthcare, banking and finance, among many others. If you are looking for a key ally for content control and digital reputation, ECO is your solution!