Questions for ‘Computers can translate languages, but first they have to learn’
To accompany feature “Computers can translate languages, but first they have to learn”
SCIENCE
Before Reading:
1. Imagine that you’re traveling in a country where no one speaks English (or any other language you know). How would you communicate?
2. Have you ever used a website or app to translate a phrase, sentence or website from one language to another? (If you haven’t, try out Google Translate or another site now.) What did you think of the results? Were they good? Why or why not?
During Reading:
1. What is the difference between machine translation and Google Translate?
2. Why is it difficult to translate words like “honeymoon” from one language to another?
3. What are the two ways that a machine translation system processes words?
4. Why is it important to use good grammar? And why is it OK to sometimes break the rules?
5. Why is it more difficult to teach a computer to translate Somali than Spanish?
6. Why is it important that a machine translation system be trained with the right type of books and not just any book?
7. Why does a machine-translation system need to grow and evolve?
8. Why should you never type made-up words into Google Translate or any other machine translation system?
9. How are neural machine-translation systems different from earlier systems?
After Reading:
1. Would you use a machine translation system to translate text that contained important medical or financial information into another language? Why or why not? Are there languages for which this might be a better option? Use evidence from the story to explain your answer.
2. Imagine that you are training a new machine-translation system. Which type of system would you use? What kind of information would you feed it? What extra features might you like it to have?
SOCIAL STUDIES
1. Somalia is a nation in Africa. Locate it on a map. Research its recent history. How might the country’s current political situation influence the books being written in that country? And how would that affect machine-translation systems?