Machine Translations in E-Discovery: Bueno o Malo?
By
Sierra M. Palmer, Esq.
Schwartz Semerdjian Cauley & Moot LLP
Published: 02.01.2019
Language barriers are arising more frequently in litigation as the world of technology expands: parties extend across geographic borders or originate from places where English is not their first language. As of 2018, he legal sector is said to be an area “most in demand” for translators[1], which can make discovery, trial, and mediation more time consuming, more complicated, and more expensive. But what are the alternatives? I think you can delete an extra space here Some say Machine Translation Technology is the wave of the future in breaking down legal language barriers, while others warn to steer clear.
Machine translation has its origins in the beginning of the Cold War and was created to translate Russian sentences into English for the first time in history. [2] However, this newly developed technology was extremely limited at the time, as it was only able to translate roughly 250 commonly-used words in Russian. [3] Since then, there have been phenomenal developments in machine translation, allowing anyone to translate any text or speech from one language to another in a matter of minutes, even seconds. So, that begs the question, what is the hesitation in using this machine translation in the legal field?
In 2017, a study was conducted in Korea to determine a human translator’s accuracy and speed versus a machine’s.[4] The results showed that the machine took only minutes to translate the mass amount of text, whereas the humans took hours.[5] However, the highest-scoring machine translator only translated 47% of the text correctly, whereas the highest-scoring human was 82% accurate.[6]
In the legal realm, it can be argued that language is everything. From the interpretation of the Constitution and legal statutes to legal jargon, idioms, and terminology commonly used among different divisions of the legal community, every good lawyer knows that one wrong or missing word can completely change the outcome of a case. This means that consequences of a mistranslation in a legal context can be severe, resulting in inaccurate testimony, wrongful convictions, erroneous discovery responses, or miscommunicated terms of arbitration. The Project Manager for a Translation and Interpretation Software company in the UK explains, “I don’t think machine translation would work for legal services, as documents contain personal details and case descriptions which make them different to each other. The complexity of legal terminology is very challenging for any machine translation tool unless it is specifically designed for that kind of text.”[7]
In State v. Pacheco, an issue arose with an interpreter’s authorization to be in a jury room during deliberations to assist non-English speaking jurors.[8] In reviewing Court Interpretation Standards for jurors and prospective jurors, the court favored certified, human interpreters and outlined the following standard for machine translations:
A number of services are available on the Internet and elsewhere that provide free or low-cost translation of written materials from English into a number of other languages. Because machine translation may not be accurate, courts should not use machine translation for written materials that are to be used in formal court proceedings, such as jury instructions or documentary exhibits. Although courts may consider using machine translation for other informational and local orientation materials submitted to jurors and prospective jurors, all courts are cautioned against relying exclusively on machine translation without human verification of the accuracy of a machine translation.[9]
On the other hand, several litigators attest to the fact that this machine translation has sped up the discovery process and lowered the cost of litigation considerably in recent months. With discovery being the most expensive aspect of civil litigation these days, cost is a persuasive factor when it comes to choosing human versus machine. Firms are now receiving translated documents sooner than they would waiting on human translation, and they are being charged a fraction of the price that human translators would cost to do the work. Most machine translation technology services charge attorneys translation fees either by the word, the page, or the document.[10] Several human translators charge by the amount of time it takes to translate the document, or they may charge a flat fee.[11] As discussed above, it naturally takes humans much longer to correctly interpret legal documentation than it does a computer, and the prices accurately reflect that extra time spent. Human translation services also calculate prices based on a variety of factors that machine translators do not, such as complexity of the subject matter, language combinations, turnaround time required, and formatting requirements—all of which add dollar signs to an attorney’s translation costs.
In addition to cost benefits, many proponents of the machine technology argue that the industry has developed a computerized technique of translation that offers cutting edge accuracy. In its simplest form, machine translation used to work by substituting each word in the sentence with the equivalent word in the target language.[12] Now, the technique used is based on a type of in-depth machine learning that is said to create a more sophisticated analysis of the language by viewing the words in a more abstract way, using algorithms, characters, syllables, and other learning techniques to better interpret the linguistics.[13]
As an added bonus in using machine translators, most of the interpretation technology can be integrated with a firm’s current document review platform, consolidating document review and document translation into one fell swoop.[14]
A combination of these two approaches is slowly emerging, marrying the concepts of accuracy and efficiency. This is called Post Editing Machine Translation (PEMT), where a machine translation is then given to an experienced translator who proofreads and corrects the automated translation in less time than it would take the translator to translate the document themselves.[15] Until machine translation technology enters into its optimal stages of accuracy, this may serve as a happy medium for firms who are looking for a cost-efficient solution to language barriers in litigation. In the meantime, both Courts and experts warn: although it may be cost effective, proceed with machine translations at your own risk in litigation.
[1] Andrew Tattersall, How will machine translation impact the legal translation sector DA Languages (2018), http://www.dalanguages.co.uk/legal-machine-translation/ (last visited Dec 7, 2018).
[2] Ilya Pestov, A history of machine translation from the Cold War to deep learningfreeCodeCamp.org (2018), https://medium.freecodecamp.org/a-history-of-machine-translation-from-the-cold-war-to-deep-learning-f1d335ce8b5 (last visited Dec 7, 2018).
[3] Id.
[4] Alison Kroulek, A Translation Showdown: Human Translation vs Machine Translation K International (2018), https://www.k-international.com/blog/human-translation-vs-machine-translation-contest/ (last visited Dec 7, 2018).
[5] Id.
[6] Id.
[7] Andrew Tattersall, How will machine translation impact the legal translation sector DA Languages (2018), http://www.dalanguages.co.uk/legal-machine-translation/ (last visited Dec 7, 2018).
[8] State v. Pacheco (2007-NMSC-009) ¶ 141 N.M. 340, 341.
[9] Id. at 354.
[10] Julie Anne Halter & Lori Steidl, Arbitration World 36th Edition July 2018K&L Gates LLP, http://www.klgates.com/ePubs/Arb-World-36th-2018/?page=28 (last visited Dec 7, 2018).
[11] Vic Marcus, Translation Pricing – How does it work?NWI Global(2018), https://www.nwiglobal.com/blog/translation-pricing-how-does-it-work/ (last visited Dec 7, 2018).
[12] Jon Kerry-Tyerman, Why Machine Translation Matters in Ediscovery The Everlaw Blog(2018), https://blog.everlaw.com/2018/01/10/machine-translation-ediscovery/ (last visited Dec 7, 2018).
[13] Id.
[14] Press Release, Iconic launches new, secure eDiscovery translation solution Slator(2017), https://slator.com/press-releases/iconic-launches-new-secure-ediscovery-translation-solution/ (last visited Dec 7, 2018).
[15] Andrew Tattersall, How will machine translation impact the legal translation sector DA Languages (2018), http://www.dalanguages.co.uk/legal-machine-translation/ (last visited Dec 7, 2018).