Why outsource machine translation directly to linguists with best-of-breed tools?

When clients directly input data into a machine translation (MT) system like DeepL, they receive raw, unedited translations. While MT has advanced significantly, it still does not fully grasp nuances, context, or specific industry terminology, leading to potential inaccuracies or inconsistencies.

In previous posts we have highlighted the various errors that can occur in machine-translated legal texts as well as the potential pitfalls for post-editors. In this article, we’ll take a closer look at why it’s much more rewarding and cost-effective, in both the short and long term, to delegate these industry-specific tasks to professional linguists.

What are the benefits of outsourcing these tasks directly to a trusted language service provider (LSP)?

There are several advantages to using an LSP that uses MT in conjunction with professional translation and post-editing services:

1. Quality assurance: LSPs employ professional translators who review and refine MT output to ensure accuracy, cultural appropriateness and adherence to industry-specific terminology.
2. Consistency: LSPs use translation memories (TMs) and terminology databases to ensure consistent use of language across all documents.
3. Efficiency: LSPs can handle large volumes of text more efficiently, using a combination of MT and human expertise to speed up the translation process without sacrificing quality.
4. Customisation: LSPs tailor translations to their client’s specific needs, incorporating brand guidelines and preferred terminology into the final product.

How does the machine translation and post-editing process work when an LSP using professional translation tools is involved?

When an LSP is involved in the machine translation and post-editing process, using computer-assisted translation tools such as translation memory (TM) systems, the workflow typically follows these steps:

1. Initial machine translation: The text is first translated using a machine translation (MT) engine such as DeepL Pro. The MT engine provides a raw translation, which is the first draft.
2. Human post-editing: Professional translators review the MT output. They correct errors, improve fluency and ensure that the translation accurately conveys the intended meaning and tone. Post-editors check that the translation conforms to the client’s style and terminology guidelines.

As translators work on the MT output, their changes are stored in the translation memory (TM). The TM is continually updated with new, high-quality translations that can be reused in future projects.

Last but not least: When a machine-translated text is post-edited in a TM environment, the TM system suggests previously translated segments for the linguist to review and edit if necessary. If the same or similar segments appear later in the text, the TM system will first check for existing translations before consulting the machine translation engine. Because the edited segment has been stored in the TM, it is retrieved and suggested for the repeated segment, ensuring consistency and reliability of the translated data.

The significance of terminology management

Terminology management ensures quality and consistency in machine translation by maintaining uniformity and accuracy of terms, enhancing efficiency, enabling better quality control, and preserving brand identity across languages. A trusted LSP can play a critical role in solving the well-known problems of incorrect, ambiguous or inconsistent terminology by helping the client build repositories of client-specific terms and managing the terminology for them.

1. Increased efficiency and productivity: Over time, clients will benefit from increased efficiency and productivity as LSPs becomes more familiar with the client’s legal terminology and preferences through the TM system. This familiarity will streamline the translation process, resulting in faster turnaround times and potentially lower costs for future projects.
2. Scalability: As the LSP’s TMs and terminology databases grow, they can handle larger and more complex projects more efficiently.
3. Continuous improvement: The feedback loop between human translators and MT engines ensures continuous improvement in translation quality. LSPs can also fine-tune MT engines based on their specific client needs and preferences.
4. Collaboration and feedback: By engaging in collaborative discussions and incorporating client input, the LSP can continually refine and update client-specific terminology repositories, ensuring continuous improvement in translation quality and consistency.
5. Training and guidance: By empowering clients with the knowledge and tools to maintain their terminology repositories, the LSP can help prevent false, ambiguous, or inconsistent terminology in future projects.

Clients should rely on trusted LSPs for machine translation instead of using public AI tools, saving time for their core business activities.

Immediate benefits

By combining MT with human post-editing, the final translation is accurate, polished, contextually appropriate and, most importantly, meets the quality requirements agreed with the client.

In addition, the immediate benefits of machine translation and post-editing of client content within a translation memory system include workflow efficiency, cost savings and improved consistency and accuracy. Faster turnaround times and cost effectiveness (reduction in human translation time) as the benefits of machine translation are still part of the package.

Long-term benefits

In the long term, clients will also benefit from using this process rather than having to manually enter individual texts into a machine translation tool.

By consistently using a TM system to process and post-edit texts, LSPs can build up a comprehensive database of accurately translated segments that are specific to the client’s terminology and style. This leads to improved quality and consistency in future translations, as LSPs can refer back to the stored translations to ensure accuracy and adherence to the client’s preferences.

A word of warning with machine translation

Despite the benefits of using a translation memory system to edit and revise previously machine-translated texts, clients may not receive the same level of quality as if the text had been translated from scratch by a human linguist.

This is because machine translations are not always perfect and may require significant editing and revision to ensure accuracy and precision, which can affect the overall quality and fluency of the final output.

In any case, for critical content, it is always advisable to consult with a trusted LSP who can advise on the best and most efficient way to handle the content and avoid unnecessary risks.

Are you unsure whether your texts are suitable for machine translation that meets your quality requirements or whether you would be better off having them processed by specialists using other methods? Get in touch with us. We will be happy to advise you on the subject of machine translation post-editing.