Artificial Intelligence

Machine translation engines already translate more than all translators in the world per day! Machine translation is meant to help translators and not replace them. Investment in AI translation is necessary to help translators work faster and deliver better results. AI can also streamline communication between international teams when integrated into intranet software or group messaging apps.

Relying on the power of Deep Learning, a subset of Machine Learning (ML) in Artificial Intelligence (AI), Neural Machine Translation (NMT) is an approach to machine translation that uses an artificial neural network to predict the likelihood of a sequence of words, typically modeling entire sentences in a single integrated model, and also able to take the context into account. A neural network is a system that can be trained to recognize patterns in data and nuances that are both language-specific and domain-specific. As a result, the translated output sounds more natural and uses better grammatical structures, ranking NMT higher than its predecessors, such as rule-based machine translation and statistical machine translation.

Gisting is the use of AI to translate content in order to get an understanding of the original content. To do this, translation teams don’t need to go through the industry’s standard translate, edit, proofread process, but configure, run, and export the machine translated output. In other words, the translation doesn’t have to be perfect. In such case, clients only need to understand the general meaning of the source document.

Clients choose to use raw, unedited Neural Machine Translation output for internal use mostly, since it works quite well to communicate the primary message. Gisting is an effective way to translate several types of content, such as social media posts, internal communication, emails and letters, contracts, RFPs and other forms. Gisting also allows clients to decide whether to have the content fully translated. By using unedited Neural Machine Translation output, you can save more than 50% of what professional human translation would cost. Choosing raw machine translation is also the fastest way to get the content translated.

Gisting isn’t a solution when you need a quality translation, especially when the content you’re translating will be direct business facing. In these situations, we advise clients to pair NMT with human post-editing service or to select our standard or advanced translation packages.

A key success factor of NMT engines is their accuracy level which is constantly improving. While they produce translations that sound very natural and human-like, they are not error-free, as Neural Machine Translation is a fairly new and still-evolving technology. It is important that the machine translation output is reviewed by a human. Post-editing is the step needed to fix errors in raw machine translated output. Post-editing is a process in which humans modify machine-generated translation to achieve an acceptable final product. Neural Machine Translation with Post-editing is a translation service that is becoming more and more popular. By using Neural Machine Translation with Post-editing output, you can save more than 30% of what professional human translation would cost.

Coupled with a custom NMT engine, the NMT+PE service works best for certain industries. Using client’s translation memories, glossaries and other relevant multilingual data, NMT engines could be trained for a specific industry domain. The AI will also improve after editing previous mistakes. As a comparison, machine translation output produced from general translation engines is less accurate and often requires re-translating the whole sentence instead of post-editing it. If you would like to know what translation process is most suitable for your needs, please feel free to contact us. We will be happy to discuss it.