2018年5月21日月曜日

Attended Singularity Salon #26

It was about Game AI this time.

The instructor was Yoichiro Miyake (Lead AI researcher of Square Enix technology promotion department)

By the way, I don't play games at all, so I was not interested in the game part. However, I participated because I was interested in AI. The talk was extremely interesting even for people who don't know about games.

Apparently in the game industry, there is a path where a student researches about game logic > get a job at game company. I hope there will be similar path made in the translation industry as well.

The most impressionable word was "Games can no longer be created by humans alone (needs AI)". I felt that the day where the same can be said for translation would come soon.


Above is translation of an article "#26シンギュラリティサロンに参加しました。" dated December 18, 2017
Translation by Hiroko Matsuda

2018年5月19日土曜日

What is Adaptive Machine Translation (AMT)?

I'll write about the difference between Neural Machine Translation (NMT) and Adaptive Machine Translation (AMT).

In NMT, when machine translation is executed on a certain text, you cannot control the term translation. Something like this:

English (source)
.......determine............................................................................................... .....................................determines................................................................................determination................................................................................. .....................................................................determined.....................

Japanese (target)
.......判別する............................................................................................... .....................................判断する..................................................................判定する................................................................................. .....................................................................判別する....................

In this manner, there is no uniformity among target translation terms (unable to enforce uniformity).

In AMT:
English (source)
.......determine............................................................................................... .....................................determines................................................................................determination................................................................................. .....................................................................determined.....................

Japanese (target)
.......判別する............................................................................................... .....................................判別する..................................................................判別する................................................................................. .....................................................................判別する....................


When the first "determine" is translated and confirmed with 「判別する」,
the machine translation learns that term and will display the term as「判別する」in the subsequent segments.

In terms of work flow, machine translation is not executed on the entire text in the AMT. The machine translation is executed segment by segment.

Therefore, although the term "post edit" is used on NMT, when AMT is being used, it is an actual "translation" (there is no concept of post editing on the latter).

*We are currently in the process of verifying AMT. If various problems can be solved, we may implement AMT in the first half of 2018.
*Since the name includes "adaptive", I think 「適応性機械翻訳」is correct. However, considering its function and contents,  the name「学習型機械翻訳」(Learning type machine translation) is not wrong.



Above is translation of an article "Adaptive Machine Translation (AMT: 学習型機械翻訳)とは?" dated December 17, 2017
Translation by Hiroko Matsuda

2018年5月18日金曜日

70% is enough

We can get a lot of work if we are satisfied with 70%. 80% is not good. 90% is also not good. 70% is just right. In this way, work will carry on smoothly and stress free.

For example, I don't really elaborate over and over again on blog articles. There are probably many misspellings (sometimes someone will point out), but I can easily correct them. There can be no mistakes on translation jobs, so I aim for 100%. However, there are many tasks where 70% is enough.

I think it's better to focus 100% on jobs that don't allow any mistakes to be made and jobs that only require 70% concentration can be a little sloppy. It's better in terms of improving productivity if we balance our concentration on our work.


Above is translation of an article "70%で良しとする" dated December 02, 2017
Translation by Hiroko Matsuda

2018年5月17日木曜日

Books I've read: Eating style taught by doctors -The ultimate textbook-68 Medically correct ways to eat after seeing 200K patients (医者が教える食事術 最強の教科書ー20万人を診てわかった医学的に正しい食べ方68)

This book is about eating. I'm writing about this because eating right leads to better work performance.

Book: 医者が教える食事術 最強の教科書ーー20万人を診てわかった医学的に正しい食べ方68
Author: Zenji Makita
Publisher: DIAMOND, Inc.

At the beginning of the book, it mentioned the following:
1. There's no relation between calorie and obesity.
2. Dieting has little effect on your cholestrerol level.
3. Protein and amino acids destroy your liver.

Next, the following was written as a medical evidence.
  • The only cause of fat is sugar (carbohydrates). You won't gain fat from eating meat.
  • Eating fat doesn't make you fat.
  • You should exercise right after your meal to prevent your blood sugar level from increasing. If you're too full to exercise, you're eating too much.
What was interesting is that "true health cannot be gained by "temporary effort). If you are a reasonable business person, you understand that it is far more beneficial for you to focus on limiting carbohydrate intake rather than trying to maintain hard-earned muscles. Walking or going up and down the stairs for 20 minutes can be your exercise." Apparently, if you're too full to move after lunch, that means you've eaten too much. It was my first time reading about how it's actually good to exercise immediately after meals. I'll give it a try.

The book also mentioned that "if you can stabilize your blood sugar level, your daily performance will improve". This is most likely important for work productivity.

There are many methods such as changing how you work and automation by software implementation. However, I want to try changing my diet to achieve work efficiency.




2018年5月15日火曜日

マッチ率0%は廃止された

機械翻訳とか、MTPEとかありますが、あまり難しく考える必要はないと思います。マッチ率の0%が廃止されて、最低でも70%から始まると考えればそれで十分ではないでしょうか。PEに特別のテクニックは不要です。70%から翻訳を始めると考えればいいと思います。こう考えるとずいぶん気が楽になるのではないでしょうか。0%から始めるのと70%から始めるのでは大きな違いが精神的にもあると思います。いかがでしょうか。

2018年5月12日土曜日

翻訳と人工知能とについて専門家にちょっと聞いてきました

グランフロント大阪の人工知能の開発会社の方とちょっと20分ほどナレッジサロンで話しました。やはり翻訳会社は、同じようなことを考えているようでした。人工知能の利用方法に関しても知らなかったことを聞かせていただきました。機械翻訳と人工知能とを組み合わせれば面白いことができそうでした。コラボレーションは、異業種と行うべきですね。

2018年5月9日水曜日

スピードアップにつながるか?

野球でたとえるなら、一塁は、走り抜けた方が速いか、ヘッドスライディングをした方が速いかという議論と同じではないかと思います。

長らく走り抜けた方が速いと言われていましたが、ある大学教授の方が調査したところヘッドスライディングした方が速いことが判明したそうです。

一塁到達、頭からの方が速い 立命大分析、野球の定説覆す
京都新聞、2018年3月20日の記事

調査方法によっても結果は異なると思いますが。

僕は、たとえ大量のポストエディットの必要はあっても機械翻訳を使わないより機械翻訳の方が速いと考えている派閥です。みなさんはいかがでしょうか?