11-731 Machine Translation: Questions and Clarifications

11-731 Machine Translation: Homework 1 - Questions and Clarifications

Question

Does "Show the correct translation of your example into English, and also give a literal gloss translation of your example." actually mean "Show the correct translation of your SENTENCE CONTAINING THE EXAMPLE into English, and also give a literal gloss translation of your example." ?

Also, you recommend that we look for real data, does it mean that our sentences should come from the web or that the examples should come from the web (e.g. for an ambiguous word with 2 meanings, should we find sentences on the Web containing this word or should we give how frequently one sense is used?)

Answer

Let me try to clarify. "Examples" refers to the unit that is representative of the particular class of translation divergence. For example, in lexical differences, the "example" would be a word that has multiple meanings and translations. For structural divergences, the "example" would be a larger unit, such as a noun phrase. For these larger units, I asked you to give a "gloss" - a word by word translation of the unit, that would clarify the divergence, in contrast with the correct translation into English of that unit.

Regardless of the unit size of these examples, they naturally occur (in the source language) within complete sentences. I suggested looking in "real data" in the source language - i.e. at online articles from Le Monde - in search of sentences that contain examples of the type you are looking for. If you find such sentences, show the entire sentence and clearly identify the divergence unit within it. It would be best if you could then give the correct translation of the entire sentence, again explicitly identifying the part that corresponds to the divergence example.

If you are unable to find "naturally occurring" sentences on the web, it is also OK to come up with an example on your own. You can then search on the web for sentences that contain your example. If you do things this way, I asked that you also give some indication of whether your example is a common or uncommon word/phrase/construction in the source language. How many hits did you get? Are they all matches of the example with the intended meaning? Comment on that...

- Alon