Google Translate announced earlier that more than 200 million people actively use the free online translation solution monthly. Given its wide variety of language pairs and instant output, the web giant’s translation service has been widely used on different occasions. It may be a perfect language solution for general documents, yet, it could be problematic for the legal industry.
Confidentiality issues
Google Translate uses a statistical approach to build an online database for translations which are produced by humans and are available online. Even with the term of service regarding confidentiality issue, a data breach is a concern that could not be neglected. Norwegian news agency NRK uncovered a massive privacy breach of Translate.com website in 2017. Users found their confidential data exposed on the internet after using the free online machine translation service powered by Microsoft Translator. Tones of sensitive information is freely accessible, ranging from personal email exchange, late payment notices to staff performance report of a global investment bank. The breach highlights the need for the language service provider to ensure that customer information is kept confidential, especially for sensitive legal documents. Elite Asia is committed to protecting customers’ confidential information upon signing a Non-Disclosure Agreement, putting privacy as the first priority.
Mix of languages
Latin and French words are still common in modern legal writings, such as ‘oyer et terminer’, meaning to hear and determine; ‘de facto’ meaning in reality. Throughout a legal document, it is not uncommon to see a few Latin and French terms embedded in one source language. Google Translate is now supporting 109 languages but it detects and translates texts from one single language to another. Mixture of different languages in one sentence may lead to awkward translation or serious error which is unacceptable for any project, let alone legal binding documents.
Long sentence structure
Google Translate’s neural machine translation system performs deep learning and learns from millions of examples to provide the most relevant translation. However, there are insufficient number of training corpora containing long sentences. Besides, considering the nature of encoding a variable-length sentence into a fixed-size vector representation, the neural network may fail to encode all the important details in a long sentence.
Unideal for uncommon language pair
The giant translation tool analyses enormous amounts of data based on previous translation and information and tries to deliver accurate result. The amount of available data is a major factor determining the quality of translation, leading to quality inconsistency of translation especially for some data-poor languages. Although the system uninterruptedly analyses different content, it still takes some time to incorporate new terms into the database. Facing idiomatic expressions and slangs words, commonly seen in testimony, which are constantly changing in a vivid real life language system, the online tool may fail to capture the real meaning of the sentence.
The tech giant puts a great effort into improving quality and accuracy of translation by implementing different new technologies and techniques. Yet, the cutting-edge algorithm model underpinning Google Translate still falls prey to different errors. It can be a treasure for causal usage, but rather risky and unreliable for legal documents.
