Difference between revisions of "Курсы машинного перевода для языков России/Session 2"

From Apertium
Jump to navigation Jump to search
Line 28: Line 28:
 
They may inflect in only one position, be it:
 
They may inflect in only one position, be it:
   
* At the end, ''vierkante meter, vierkante meter'''s'''''{{slc|nl}} or
+
* At the end, ''pazartesi gün, pazartesi gün'''ler'''''{{slc|tr}} or
 
* In the middle ''hôtel de ville, hôtel'''s''' de ville''{{slc|fr}}
 
* In the middle ''hôtel de ville, hôtel'''s''' de ville''{{slc|fr}}
   

Revision as of 20:22, 21 December 2011

This session aims to give an introduction to some more advanced topics in the Apertium dictionaries. We have seen in session 1 how to make simple dictionary entries, and looked at paradigms. In this session we'll make entries for multiwords.

Theory

When talking about multiwords (or multiword expressions, MWEs), there are a few aspects to take into account.

Compositionality

TODO; find examples

A multi-word expression is compositional when its meaning can be deduced from the meanings of the words that comprise it. However, this is not to say that it will necessarily be compositional in translation. For example:

  • Телефонная будка(ru) is compositional, the meaning can be inferred from the meanings of the two words.
  • Juego de etiquetas(es) is a multiword expression that is compositional, but the sense that each of the content words has is not the most frequent one.
  • Итальянская забастовка(ru) is partially non-compositional, the meaning cannot be deduced solely from the meanings of the constituent words.
  • Imeti mačka(sl) is non-compositional, the meaning cannot be deduced from the meanings of the constituent words.

This is important for translation as a compositional multiword might be possible to treat just as two simple words, while a completely non-compositional one has to be fully listed.

Variability

TODO; find examples

Multiwords may allow variation, either for inflection, e.g. French vaut la peine, valait la peine, or by allowing modifying words, Spanish: tener ganas de, tener muchas ganas de. Multiword verbs may allow passivisation or not.

They may inflect in only one position, be it:

  • At the end, pazartesi gün, pazartesi günler(tr) or
  • In the middle hôtel de ville, hôtels de ville(fr)

Or on both sides:

  • телефонная будка, телефонной будки (ru) and
  • estoy volado, estáis voladas(es).

Contiguity

TODO; find examples

Along with internal variation, multiwords may also be contiguous or discontiguous. Contiguous means that the words follow each other, separated by spaces. Most of the above examples are contiguous multiwords. They can also be discontiguous, like phrasal verbs in Swedish and Dutch, examples might be: lopen uit de hand, äta upp and tilføje:

  • Twee jaar later loopt de situatie in het land onder de indruk van massabetogingen uit de hand.(nl)
  • Alla stygga barnen åt tigern upp.(sv)
  • Føje et ord eller en sætning til den brugerdefinerede ordbog.(da)

Practice