How computers translate human language - Ioannis Papachimonas 13.03.

2019
https://ed.ted.com/editor/1198470

Is a universal translator possible in real life? We already have many programs that claim to be
able to take a word, sentence, or entire book in one language and translate it into almost any
other. The reality, however, is a bit more complicated. Ioannis Papachimonas shows how
these machine translators work, and explains why they often get a bit mixed up.

1. What do you think are the pros and cons of a "Statistical Machine Translation"?
2. Why would semantics play an important role in a machine translator?
3. Using your own innate knowledge of your native language, give one or two examples of an
irregularity or exception that could cause problems for a machine translator.
4. Which of the following is a method used in designing machine translators?
a. Rule-based translation
b. Audiovisual translation
c. Machine-assisted translation
5. Why is it difficult to construct a machine that translates perfectly?
a. It is difficult for computers to understand all the exceptions, irregularities and shades of
meaning that come instinctively to humans
b. Machines lack the computing power to process all the linguistic data
c. We can't successfully construct a set of rules for a human language
6. "Statistical Machine Translation" is based:
a. On a set of linguistic rules
b. On an existing database of translated documents
c. On a set of mathematical formulas
7. Using which part of theoretical linguistics, how would a machine translator distinguish "child"
from "children"?
a. Syntax
b. Morphology
c. Phonology
8. A machine translator rearranges word order in the target language. What is one of the main
problems that this could cause?
a. It causes no problems at all
b. The generated text wouldn't be aesthetically pleasing
c. It could change the meaning in the target language
d. It could change the meaning in the target language
e. The generated text wouldn't be aesthetically pleasing
f. It causes no problems at all