Definition of MT - 30 Years Ago Back in 1993
Here is a copy of a vintage year-book dating back to 30 years ago:
imidas published by Shueisha Inc. （集英社出版『イミダス』).
Many Japanese publishers used to print a (hardcopy!) encyclopaedic anthology of the trend words of the year.
It was fun to find the entry for “machine translation” in this antique copy. On this International Translation Day, let me show you what I found there.
機械翻訳システム
コンピュータを用いて翻訳を自動化ないし支援するシステムのこと。（中略）人工知能の手法を使うことにより翻訳システムそのものの質が向上していることは確かである。そして、もっとも重要なことは辞書の電子化であり、そのために電子化辞書プロジェクトが八五年よりスタートしている。
(If you will allow me to translate the last section,
it reads: “What is most important is the digitisation of dictionaries. To this end, the electronic dictionary project was launched in 1985.”)
This was long before machine learning, deep learning and neural machine translation (#NMT). Although the concept of MT was already in vogue in those days, the efforts described here sound very slow and tedious.
We have come a long way since then.
Out of curiosity, I did some research online to see if the imidas is still available. As I suspected, it ceased to exist in its printed form in 2006. As of 2008, its contents were transferred to a web-based service,
The site is packed with useful glossaries and blog-like articles. It is well worth taking a closer look at.
Amidst an overall slump in the publishing world as it fails to woo digital native consumers, traditional publishers are having to fight for their survival by repositioning themselves in order to regain their appeal.
It may be the case that the same paradigm shift is also taking place in our own industry right now if MT proves itself to be a true destructive technology.
After all, that’s why I am writing all this here today: in order to allow my voice to be heard. When human translators post on social media, you can get a sense of who they are. But the same cannot be said for MT engines with black box algorithms that use big data of unknown authorship as their feedstock.
#InternationalTranslationDay #HT #MT