Machine Translation (MT) is a powerful Localization Industry solution that has developed significantly in recent years. With the advent of more advanced neural models, MT engines have become more sophisticated than ever, capable of processing vast amounts of data with ever increasing accuracy. MT is particularly useful when large volumes of content need to be translated in a short amount of time, or when a more traditional translation workflow may be cost-prohibitive or Machine Translation is not a suitable solution for specific content like marketing or advertising.
On its own, raw machine translation isn’t always appropriate. We’d never recommend it for high-visibility content such as marketing materials or ad copy, since it can’t translate the nuance and subtlety of language the way a professional linguist could, but with the right workflow and experienced human post-editors to work on content, the results can be as perfect as you would expect from exclusively human-translated workflows. Raw translations from MT may be suitable for such content as blogs or Knowledge Based Articles where a precise translation is not required.
Custom Workflows and MT Engines
When we implement a machine translation solution at Dynamic, we tailor the entire process to our client’s needs, and we guarantee 100% confidentiality and security for their data, beginning with an analysis to help them determine whether MT is even appropriate for their content. If so, we analyze and discuss which type of MT would best suit their needs, and whether it would make sense to include a human post-editor and legacy translations into the workflow.
We develop custom neural and statistical machine translation solutions which utilize custom-trained “smart” engines that can learn from human-generated translation, allowing the engines to become increasingly efficient. The MT engines provide higher quality output as they receive more input, whether via previously translated content, newly edited content from our translation professionals, or via input from the client’s team.
The most critical component for a high-quality automated translation is a post-editing and review process handled by professional linguists. This ensures that a human linguistic expert will enhance the quality of translated content by correcting grammar, nuance, tone, calls to action, terminology, and punctuation. In the case where clients have their own internal post-editors, we work with clients on Best Practices and Style Guides to compliment the MT Workflow.
By using a certified translator to handle machine translation post-editing, content is reviewed against the source, and edited where appropriate to bring the content closer to the intended meaning. As the linguist provides feedback, the MT engine becomes more accurate, reducing the post-editing “gap” and making future work less expensive.
Secure Self-Service Solution and Pricing
We also create custom “self-service” solutions with incredibly low-cost machine translated content that client teams can log into and utilize, where the results are not just generic machine translation from a publicly available engine, but rather pre-populated with their approved content in a secure and confidential environment. Whether it makes more sense to have a quarterly or annual subscription, or instead to have a per-word fee with a minimum quarterly volume, the team at Dynamic is happy to customize the solution to best fit any specific need or budget.
For every MT-based project, our in-house MT Engineers will analyze the source content against the pre-existing language(s) and client-specific translation memories we manage in-house. Any applicable Translation Memory content, glossaries, and designated non-translatable content provided will assist in training the custom MT engine to more accurately choose the correct translation for each phrase. The MT Engineer may also strengthen the engine with corpora (related previously translated content).
Meanwhile, for content designated for human post-editing, a point of contact at Dynamic will provide glossary and reference content to each of our linguists involved in the process, thereby increasing expediency, efficiency, and terminology consistency. At the completion of every project, all final translations will be integrated back into the Translation Memory files to leverage against future translation projects. Translation Memory content can also be used to retrain the base engine on a regular basis so that the predictive neural or statistical MT will be increasingly accurate for the client’s content.
Dynamic Language partners directly with over a dozen industry-leading machine translation developers across multiple platforms, but we can work with virtually any MT engine available. If you’re interested in learning more about how we can develop a custom Machine Translation process for you, please contact us. We‘d be more than happy to help you find the best solution for your organization.