Back to Blog

The Evolution of Translation Services in eDiscovery

March 29, 2024
Translation service from start to end in eDiscovery

In a global economy, lawsuits and government investigations frequently involve foreign language documents. Case teams can face thousands or even millions of documents written in multiple different languages. Lawyers in eDiscovery and litigation support professionals must quickly arrange for the translation of these documents so the case review can proceed.

With the advent of eDiscovery, many translation services emerged to do data translation. These businesses offered human translators with some legal knowledge. The manual translation process proved to be very costly and time-consuming. Review and production could be significantly delayed when large volumes of foreign language documents had to be translated.

Translation services have evolved over the years, with a tech revolution leading the way. However, a few things that have remained constant are the importance of translation accuracy, document chain of custody, security, and confidentiality.  Let’s take a look at how technology has revolutionized translation in Discovery and where it’s going…

On-the-Fly Translation Apps for eDiscovery

About twenty years ago, publicly available translation Apps started to arrive on the scene. Google Translate is probably the best-known. Introduced in 2006, it’s a free tool that translates text and other media from one language into another. The translation is done “on the fly” – you drag and drop your documents/text into the App and the translation immediately appears.

But for eDiscovery professionals, data security and confidentiality concerns are big red flags with on-the-fly translators like Google Translate. Uploading your documents to an app poses security concerns – do you know for sure what happens to your data after you upload to a 3rd party wesbite for translation? If a hack occurs, will your client’s sensitive information be compromised? There is also a risk that your document chain of custody could be challenged if you use an on-the-fly translator.

Early Machine Translation Apps

Early machine translation Apps were rule-based. The machine translation software analyzed the source text, word for word, using a set of rules.

Early versions of Google Translate used the statistical machine translation (SMT) method. This method leverages the most common previous translations to translate a document or file phrase by phrase.

SMT Shortcomings for eDiscovery

Since then, legal tech companies and eDiscovery service providers started offering SMT-based machine translation in eDiscovery. The tool’s website often came with a plug-in to eDiscovery platforms.

However, there are shortcomings with SMT technology for data translation in eDiscovery. With SMT, a translation is only as good as the tool’s existing reference texts. Words or phrases in eDiscovery documents that aren’t in the SMT translator’s library won’t be found or will be garbled or translated out of context. SMT translators can give you the general gist of the text. However, they can fall short when translating grammatically incorrect language, idioms,  cultural nuances, and context in language.

Anyone using SMT technology for legal data translation will want a human to review the translation results. In some instances, you may be able to limit the human review to key documents or documents from particular custodians.

From a practical perspective, your translation tool should ideally be well integrated with your discovery solution. If you use standalone solutions for data translation, you’ll have to migrate the translation results to your eDiscovery review platform. With a massive number of foreign documents, this will be time-consuming. Also, there is always a risk that data gets dropped during the migration.

Big Step Forward in Accuracy: Neural Machine Translation

Neural Machine Translation (NMT) combines deep learning techniques that mimic human neural and artificial intelligence to teach a machine to use patterns to generate a correct translation. Neural translation systems try to work similarly to the human brain, constantly looking for the right patterns and making decisions.

The NMT software is “trained” to find the most accurate translation based on examples of translations for specific text. Think of a search for patterns in the source language that match patterns in the target language.

The machine learns to recognize patterns in the source material that help it interpret the context, leading to more accurate predictions of likely word sequences. NMT technology constantly learns from feedback throughout a translation,  improving its ability to translate your foreign language documents more precisely.

Due to the ongoing learning capability, NMT is more accurate than SMT and other translation technologies. eDiscovery data translations using NMT will typically require less human editing/review. These factors are moving machine translation in eDiscovery towards the NMT model.

As of 2020, NMT could instantaneously translate texts with 60-90% accuracy. Ten years after releasing Google Translate based on SMT, Google modified the translator to use NMT. Google reported that its NMT system reduced translation errors by approximately 60% compared to its earlier SMT version. So, for now, you will still need to do some quality checks on your NMT translation results though.

It should be noted that a 2023 translator service analysis of Google Translate warned that due to its limitations, the tool shouldn’t be used for official or important business documents.

So, is the Future Generative AI (GenAI) Translation?

GenAI, still in its infancy in terms of being proven for use in legal matters, will likely turn out to be a staple for ESI translation in eDiscovery down the road. ChatGPT is the most famous example of GenAI. GenAI goes beyond translating an input into an output. This technology can create new translations because it learns from sequences of sentences. GenAI tools appear to be better at recognizing context and producing translations that more precisely present the tone and meaning of the original content.

NB – Keeping Your Data Secure and Compliant During Translation

With sensitive client information in legal matters, assessing your eDiscovery translation service provider’s data security and regulatory compliance protocols is important. You’ll want to confirm your data translation provider uses secure file transfer protocols and high-grade security protocols to protect your client data stored in their systems. Or better yet, use a provider whose translation tool is completely integrated into their Discovery solution.

Look for eDiscovery Systems with Machine Translation built in!

One of the best solutions to consider for data translation is integrating a machine translation engine into your eDiscovery platform, so you won’t have to worry about the security of your data stored at a translation service’s repository or data in transit. You’ll also eliminate the cumbersome task of moving translations done in a standalone translation tool into your review system, as discussed above. This makes for a smoother workflow. The risks of data loss or corruption on the move also go away. Stay open to change and carefully evaluate eDiscovery solutions that embed machine translation into your systems, translation services are constantly evolving!

Author

Biju Christian
With over 16 years of experience in the professional services industry, Biju brings a wealth of knowledge and expertise to his role at Knovos. As a Senior Project Manager of the professional services division at Knovos, he oversees a wide range of projects, ensuring the delivery of high-quality services to effectively meet client needs. Biju is recognized as a visionary leader, whose instrumental contributions have led to the delivery of unparalleled post-sales support by the professional services team to our esteemed clients. He skillfully navigates communication and project strategies with Knovos' diverse clientele, which includes law firms, corporate legal departments, and government agencies. With comprehensive expertise in all technical and analytical facets of the eDiscovery project lifecycle, Biju is dedicated to delivering optimal outcomes for Knovos' clients.