Introduction
In my time at MIIS, I’ve learned a lot about the translation process and the technology that goes into it. With machine translation (MT), it is often hard for a layman to understand what exactly is happening when you paste words from one language and it displays a translation right beside it. Many still don’t understand how translators haven’t been wholly replaced by machines, nor that there are many types each trained on different algorithms such as rule-based, statistical, and neural MT. However, as a computer scientist and a linguist, I learned how to stop worrying and love the machine (translation).
Neural Machine Translation (NMT) is the perfect example of how language and technology collaborate to create something entirely new and different. Described as a “black box”, NMT takes large amounts of data and eventually becomes trained to translate new data related to what it was given. With this in mind, my group, Dayna Brown (she/her/hers), Nicholas Niculescu (he/him/his), and I decided to venture into NMT with what may be the most stereotypically Floridian project we could choose: translating Disney maps from English (en-US) to Chinese (zh-CN). As the fictional LSP Lao Shu Translations, we created a pilot project to train and test a NMT to see if a custom MT engine would be beneficial and cost-effective for a company like Disney.
Here was our process:
The Proposal
Short Disclaimer
This pilot project was conducted with no formal association to Disney or other companies.
Copyright Disclaimer: under section 107 of the Copyright Act of 1976, allowance is made for “fair use” for purposes such as criticism, comment, news reporting, teaching, scholarship, education, and research. This project is a proof-of-concept, and as such does not represent nor infringe on the creator(s) in any way.
