Machine translation (MT, henceforth) is now considered reasonably accurate and has been used in public places in Japan, including stores, stations, and hospitals (Nikkei, 2019). Google Translate adopted Neural Machine Translation in November 2016 and dramatically improved its translation results (Google, 2016). In 2017, Mirai Honyaku claimed that their MT system was equivalent to a person with a TOEIC score of 900 or more when translating Japanese into English.
A symposium with around 150 attendees at a JACET (Japan Association of College English Teachers) conference held in March 2019 was significant as its three presenters agreed that it is time for foreign language teachers in Japan to think about how to live with MT and possibly make positive use of it (Baba, Garry, and Narita, 2019). Thus, the next question is what exactly can you do with MT in your classroom?
The purpose of this column is to introduce some tips that can help college students in Japan use MT effectively. I will explain how MT can be a useful tool for writing speeches in English. I also suggest that teachers encourage their students to follow what I refer to as the Golden Rules when they use MT. Finally, I will highlight how MT is more useful than regular dictionaries in some cases.
Tips, Rules, and Advantages of Using MT in Making Speech Drafts
Novice–intermediate students usually need their teachers’ support in composing their speech drafts in English. That is where MT can be an additional source of English. Some of the recommended MT applications are Google Translate (Google) and VoiceTra (NICT), both of which are free. Google Translate is known for its accuracy (Turner, 2016). VoiceTra provides “reverse translation,” which allows students to check whether their input has been translated the way it was intended.
To obtain optimal translation results, there are three things that may help students. First, try different fonts and punctuation. Hiragana, katakana, kanji, and even the use of a comma produce different results. Using kanji often helps disambiguate the input. Second, try multiple inputs. Changing verbs or nouns is especially useful. Third, include overt subjects in the input. It may sound unnatural in Japanese to have overt subjects, but it is better to have them in order to get better translation results. In what follows, x→y means “Google Translate translated x to y on October 3, 2019.”
Tips
1. Try different fonts and punctuation (hiragana, katakana, kanji, comma, etc.)
1st attempt: だいこんをほそくきったほうがいいです。 →It’s better to dig a lot.
2nd attempt: 大根を細く切ったほうがいいです。
→ It is better to cut the radish into thin pieces.
2. Try multiple inputs
1st attempt: 私はお皿を下げました。
→ I lowered the plate.
2nd attempt: 私はお皿を片付けました。
→ I put away the dishes.
3. Have overt subjects
1st attempt: 父は買い物に行きました。でも何も買いませんでした。
→ My father went shopping. But I didn’t buy anything.
2nd attempt: 父は買い物に行きました。でも彼は何も買いませんでした。
→ My father went shopping. But he didn’t buy anything.
All the translations in the 1st attempt are odd, whereas the ones in the 2nd attempt are precise. These three tips are just some of many. Students may figure out more by themselves as they get used to MT.
The next step is very important: I suggest teachers let students follow what I call the Golden Rules, as given below.
The Golden Rules
Adopt a translation result if and only if:
1. You understand it.
When you do not understand a translation result, it might be wrong or too difficult for you.
2. You can memorize it (or, at least, read it smoothly) for your speech.
Unlike teachers, MT systems do not know anything about you. You need to choose English that is appropriate for your proficiency level. If you cannot memorize a translation result or, at least, read it smoothly, it is not for you.
3. You bear responsibility for what you say.
Do not speak incomprehensible English and make excuses for it by blaming MT. You are responsible for choosing the expressions in your speech.
As you can see, speech-writing activities help students choose appropriate English for their proficiency level. When students are required to submit written work only and do not have to make speeches, they may end up choosing translation results that are beyond their proficiency level. In other words, speech-writing activities help minimize this negative effect of MT.
Finally, I would like to point out that MT can sometimes be more useful for novice–intermediate students than regular dictionaries. Three examples are given below. First, Japanese speakers find it difficult to follow the rules of gender and number agreement, even though they learn these things in middle school (Shirai, 2004, p 89). MT can provide a concrete verb form for a particular subject. Second, students find it difficult to obtain appropriate verb and object combinations. For example, for 夢を見る, Japanese-English dictionaries have numerous options where見るincludes see, watch, look, and stare, and it may be hard for novice–intermediate students to come up with good combinations, such as have a dream. Third, MT is useful in translating numbers. The population of Japan is about 一億三千万 (130,000,000). As Japanese uses a different number system than English, it is not an easy task to come up with one hundred and thirty million.
Advantages of MT
1. Subject-verb agreement
誰もその人を知りません。
→No one knows that person.
2. Verb and object combination
私は昨日の夜変な夢を見ました。
→ I had a strange dream last night.
3. Numbers
日本の人口は約一億三千万人です。
→ Japan has a population of about 130 million.
Conclusion
I have presented some tips on how to extract better translation results using MT, suggested that some drawbacks of MT are likely to be reduced if students follow the Golden Rules, and explained how MT can sometimes be more useful for novice–intermediate students than regular dictionaries. Despite these strategies, there is no doubt that students still need support from teachers to help frame the topic and improve the structure and delivery of their speech. “Nothing will be able to replace the help and guidance provided by human teachers” (Lynn, 2016), but MT can be a useful assistant.
References
Baba, T., Garry, T., and Narita, J. (2019, March 10). AI ya honyakuki ga hattatusitara gaikokugo kyouiku wa dounaruka [What happens to foreign language education when AI and MT advance], Symposium conducted at the meeting of JACET gengokyooiku ekisupo [JACET language education expo], Waseda University, Tokyo.
Google Japan. (2016, November 16). Google honyakuga sinka simasita. [Google Translate has improved]. Retrieved from https://japan.googleblog.com/2016/11/google.html
Lynn. B. (2016, December 18). How will Machine Translators change language learning? VOA Learning English. Retrieved from https://learningenglish.voanews.com/a/how-will-machine-translation-chang...
Mirai Honyaku. (2017, June 28). TOEIC 900 ten izyou no eisakubunnouryoku o motu sinsou gakusyuu ni yoru kiai honyaku enjin o ririisu [Mirai Translator TM, TOEIC 900 point or higher in translating Japanese into English is released]. Retrieved from https://miraitranslate.com/uploads/2017/06/2d5778dcdee47e4197468bc922352...
Shirai, Y. (2004). Gaikokugo gakusyuu ni seikousuru hito shinai hito daini gengo syuutokuron e no syoutai [Those who succeed in foreign language learning and those who do not, an invitation to researches of second language acquisition]. Tokyo, Japan: Iwanami Shoten.
Turner, K. (2016, October 3). Google Translate is getting really, really accurate. The Washington Post. Retrieved from https://www.washingtonpost.com/news/innovations/wp/2016/10/03/google-tra.... 0f1ee7f53184
VoiceTra. (n.d.). Retrieved from http://voicetra.nict.go.jp/en/