Detection of translation equivalents in comparable corpora Serge Sharoff Centre for Translation Stud

1

Detection of translation equivalents in
comparable corporaSerge SharoffCentre for
Translation Studies,University of Leeds
2
The problem

• Why Germans dont feel anger
• Uwe Durst, (2001) In J. Harkins, A. Wierzbicka,
  (eds.) Emotions in Crosslinguistic Perspective.
  Berlin Mouton de Gruyter. 115-148.
• Ärger in, Ärger mit, Ärger über (sich ärgern
  über)
• Anna Wierzbicka, (1998). Sadness and anger in
  Russian the non-universality of the so-called
  basic human emotions. In A. Athanasiadou and
  E. Tabakowska (eds.) Speaking of Emotions
  Conceptualisation and Expression. Mouton de
  Gruyter. 3-28.
• anger?gnev?Ärger

3
Going beyond dictionaries

• frustratingly
• First-class citizens frustrate, frustration
• Second-class citizens frustrated, frustrating
• Third-class citizens frustratingly
• The two parties were frustratingly close to an
  agreement.
• A false friend ?? ????, ??? ??? ?????????
  ????????? - ?? ?????, ????????, ??????????
  ??????????.

4
frustratingly in corpora

• Parallel corpora are too small in the BNC
  frustratingly 0.42 ipm, frustratingly
  pos"JJ" 0.24 ipm
• Comparable corpora BNCReuters (190 MW), RRC (50
  MW), http//corpus.leeds.ac.uk (concordances and
  collocation lists)
• Functions of frustratingly pos"JJ"
• What is evaluated Judgement vs. Appreciation
• J a real do-or-die enthusiasm thats
  frustratingly absent.
• A Trust is frustratingly difficult to measure.
• A Comets are frustratingly rare.
• J The parties were frustratingly close to an
  agreement.

5
Data to use

• Semantic classes manual (lists of synonyms in
  dictionaries), automatic similarity (Manning,
  Schütze, 1999)
• Collocations (MI score, T score, Loglikelihood)
• Equivalence classes manual (bilingual
  dictionaries), automatic (p. 160, Manning,
  Schütze, 1999)
• Detect stable translation equivalents close to
  an agreement

6
Translations of close to an agreement

• Collocations of words in the Russian class
  ??????? ??? ?????? (quite), ????? (almost),
  ?????????? (sufficiently), ??????? (by far),
• An additional condition close to an agreement -gt
  ?????? ? ?????????? ?????????? (??????????,
  ?????????????, ?????????, ?????, ?????????,
  ??????)
• Collocations of words in the Russian class
  ???????resultagreement ???????, ??????, ?
  ???????, ???, ????????, ???????

7
Basic points

• Translations of terms vs. general lexicon
• Use of comparable corpora is not efficient, but
  useful for training translators
• Semi-automatic methods for generating examples in
  the two languages