Machine Translation (MT) can be a daunting subject with its multiple training approaches and long list of terms. Those new to MT can easily get confused when words like Curated MT, bleu scores, corpus availability, preprocessing, and post editing are thrown around.
One such term is “gisting.” It’s brought up frequently and is a huge part of how MT is used, but what does it really mean, and what do you need to know about how it’s used?
When Should and Shouldn’t You Gist?
Simply put, gisting is the use of MT to translate foreign text to get an understanding of the original content’s meaning. To do this, the translation doesn’t need to be perfect. It just needs to be good enough to extract a general meaning of the foreign text.
Gisting is an excellent way to translate several types of content, such as internal company communications, get an approximate idea of what foreign contracts contain, review foreign RFPs and other web forms. When foreign text is encountered in any of these content types, gisting comes in handy because the translated material doesn’t need to be 100 percent accurate to be useful.
Typically, gisting isn’t enough when you need a quality translation, especially when the content you’re translating will be outward-facing. In these situations, it’s usually better to go with human translators. Gisting will often have minor to major grammatical and syntax errors that make reading awkward and the flow of the content disjointed. It often feels as if “a computer translated it,” which...of course it did! When presenting your brand or targeting engagement with your audience on key points of value, product, or message, gisting usually won’t work.
Gisting and Data Security
Many services offer open source MT solutions for users, and gisting is regularly used on an ad hoc level within a company. Google and Microsoft both offer effective gisting platforms that are effectively free to use.
Nevertheless, there are issues with using open source options to gist internal company documents. Data security is a significant part of any business’s data flow, and these open platforms are not secure. The platform providers will take your content and index it. If you were to gist a potential M&A document or new drug study findings in this way, the content would be available from that point on, in the public domain.
Gisting with Secure Machine Translation
The answer to these data security issues is secure, private MT. ULG and other LSPs can set up secure, private MT that can also be trained to your company's specific terminology, voice, tone, and style, making gisting an even more useful option for internal company content while still providing extremely cost-effective translation at a small fraction of a cent or euro per word.
Gisting is a great application for MT when used correctly for your needs--whether that’s open source Microsoft or Google, or your corporate security demands mean that a trained/curated and secure MT is a better option.
If you have any questions or want to know more about any aspect of MT, please get in contact with us and we will be happy to set up a call with our industry-leading data scientists and MT experts.