Is a bridge between human and machine translators possible? We’re already making it happen.
Roughly 5 years ago, pioneers began using the enormous computing power that’s been made available recently to explore the possibilities of using neural networks (statistical learning models that were first applied in speech and image recognition technology) to enhance existing machine translation (MT) technology, creating what’s come to be known as neural machine translation (NMT).
How does NMT work?
NMT lets machine translation engines train themselves using a trial and error process that’s similar to the way a human brain works. This process is called “deep learning” and comes from principles that have been established through the implementation of Big Data analytics. The potential of NMT is still being measured, but what’s already clear is that it nearly always improves translation quality and produces a more “human-like” output.
What difference can NMT make?
When applied properly, NMT can drive the following positive outcomes:
- Increased productivity and reduced time-to-market. On average, an NMT-based approach can increase daily translation throughput three to four times. Simply put, NMT gets content in front of your customers faster than ever before.
- Lower costs. NMT allows for the charging of lower rates, and in the long term all future projects from the same subject area will leverage off the savings achieved by applying it.
- Better consistency. Large projects require the use of multiple human translators, so achieving the same tone and control of terminology is typically more challenging than with a customized NMT solution.
The Argos approach
At Argos, our approach to NMT is all about making sure that it’s right for a client’s content. To do that, we first run a pilot project where we test both the content and the engine to determine the quality of the output generated by the engine and the amount of post-editing work needed. At this stage we also take into consideration the type of content – machine-based translation may be perfect for internal technical documents but a poor fit for external marketing materials, for example.
The first step in a pilot project is to select the appropriate NMT engine. At Argos, we follow a “technology agnostic” approach that uses multiple NMT solutions, including MT platforms that allow us to build a customized MT engine with customer content. We use several commercially available NMT engines, making sure that all translated content is safely and securely transmitted via API connections. We then run the sample content through multiple MT engines that provide good “raw MT” output for tested language pairs. If we’re provided with enough bilingual content, we can also build a customized MT engine.
The automated checks allow us to eliminate the engines that provide low quality “raw MT” for a particular type of content and language combination. Once we shorten the list of potential MT engines, we run human post-editing tests to evaluate the productivity of post-editing for predefined quality levels. This step provides detailed information that will help plan translation projects using NMT technology. Productivity levels can differ between different types of content and language combinations – there is no “one size fits all” here.
Thanks to our processes, our clients can rest assured that NMT will enhance the quality of our translations while introducing additional savings and increasing productivity.