Machine Translation Technology – Description

At Pangea, we work on Artificial Intelligence to provide machine translation technology custom developments, deployment and consultancy services to companies and institutions that require fast and professional language results handling language data. Pangea works primarily for digital, cognitive corporations, consultancy firms, manufacturing industries, translation departments within organizations, governments and institutions. We make full use of state-of-the-art computer applications to automate and accelerate language translation.

Finger pressing a key on a keyboard to request neural machine translation technology

We can help you plan and implement intelligent machine translation technology and new language automation policies with neural MT at their core to optimize your existing translation workflow. You will benefit from our past experiences and deployments at EU and US government level.

Pangea is the result of successful, tested Machine Translation Technologies + post-editing workflows implemented at Pangeanic and other organizations.

 

Machine Translation Technology

Services we offer:

 

machine translation technology as part of Artificial Intelligence, Deep Learning, Machine Learning, Problem Solving, Cybernetics

 

Consultancy on MT solutions as complementary or substituting Translation Memory systems

The majority of our consultancy services are directed to companies requesting advice on neural machine translation (NMT), hybrid machine translation or some kind of MT system integrated in their current workflow. The jewel in the crown is the use of an integrated API with which you can create and manage your own engines and offer machine translation technologies and your own output to your clients or integrate it with their Content Management Systems.

Our consulting services lean towards open source solutions or very low-cost post-editing tools. A frequent consultancy request is “how can I move away from costly CAT or TM systems to a translation software I can own and update myself, with low maintenance costs?”.

Identifying specific areas for improvement

Within our consultancy services, we will identify all areas of your organization where Machine Translation Technology can be of benefit in terms of time and saved cost. This can be as an internal translation department within an organization, analyzing whether you can host the solution internally or externally. You may find areas of inefficiency in your company or departments that translate too much or take up too many human resources.

Groups of LSPs, associations or LSPs (translation companies) with large, domain-specific account can also benefit from the development of a specific neural MT engine.

Design of customized Neural MT cycles for your organization.

There are no two identical companies offering exactly the same product. We know you don’t. Why should you rely on ready-made, out-of-the-box software or free online systems that use your data for somebody else?

Often, the translation, localization and language processes are not integrated into the overall business process. Paradoxically, no enterprise undervalues the importance of international revenue and reaching international audiences and clients.

PangeaMT will design a customized MT cycle for your organization that fits your exact needs, seldom someone else’s. The core will be a state-of-the-art neural MT engine, but you will have your own language model, preferred terminology and language turns.

Our technical department will engage with you to guide you through the data collection process, data cleaning, terminology standardization, training process, and offer training on post-editing if necessary. Hosting can be at your organization or at Pangea. Low-cost updates for fine-tuning will be scheduled at design stage.

Neural Machine Translation Technology training / testing / development and implementation

This is the core of the process, and perhaps the one clients get least involved with – however, it is the most exciting. Our programmers will make use of the data you provide to customize the engines so they reflect your preferred style and terminology. Secondly, we will also harvest data from our massive repositories and self-generate content if required to build larger and specific content for you. Thirdly, we will build language models and, finally train the engines to start thinking and translating like you want.

Just like CAT-based systems, neural machine translation systems can be updated and improved. The also get better and any specific issues can be corrected. Remember neural machine translation is not an out-of-the box standard package: it will be tailored to your organization needs. Our engineers offer training courses at programmer’s level so you can implement a sustainable, self-sufficient virtuous cycle update policy or subcontract it to Pangea for a small fee and only when required.

NMT and Human Review / Post-editing for publication

There is a clear lack of qualified translators in the world. Moreover, there is even a larger shortage of experienced human reviewers / post-editors with the will and experience to deal with high-quality neural MT. The translation profession has undergone a massive change over the last decade and machine translation technology is responsible for the change in paradigm. Humans can concentrate more on making the sentences “human” whilst leaving the hard translation work to neural networks. Such professionals are hard to come by. And yet, production increases once a translator becomes an the reviewer of a machine-generated output, with full knowledge of the source and target languages.

With Pangea, you have the choice to review or post-edit the work with your current vendor or trust our language division Pangeanic to complete the work for you. Our reviewers have been carefully selected after testing and are all native speakers of the target language. Most live in the country of origin, so they are up-to-date with current affairs and idioms. They possess an extremely good understanding of the source language. Their job is to read the original and the first version produced by our nearly-human neural machine translation technology so that it reads seamlessly, as intended, without any trace of it being software-generated.