GALE: Translation

LDC creates translations for a large volume of newswire, broadcast news and broadcast conversation transcripts, and weblog and newsgroup text. All translations are created by outsourcing to selected agencies.

All incoming data undergoes a rigid quality control procedure to ensure translation quality, and LDC staff perform manual corrections on devtest and evaluation data. Enhancement of evaluation data includes the insertion of translation variants (alternates) into the translations to represent ambiguity in the source.

LDC also collects a large volume of parallel text from existing sources.

LDC manually identifies Sentence Units (SUs) in the source text before it is sent out for translation, so all incoming translations are perfectly aligned.

Guidelines for Full-Document Translation

  GALE Arabic Translation Guidelines V2.6 - Updated 11/19/2009

  GALE Chinese Translation Guidelines V2.8 - Updated 11/19/2009

Guidelines for Single-Sentence Translation

  GALE Arabic Sentence Translation Guidelines V2 - Updated 11/19/2009

  GALE Chinese Sentence Translation Guidelines V2- Updated 1/22/2009

Guidelines for Translation QC and Adding Alternates

  GALE Arabic Alternation Guidelines V2 - Updated 11/19/2009

  GALE Chinese Alternation Guidelines V1.2 - Updated 11/19/2009