The BAF
is a corpus of English and French
bitext: it consists of pairs of English and French documents, which are
translations of one another, and whose sentences have been aligned. The
corpus was produced by researchers at the CITI, a Canadian
government research laboratory, as part of their contribution to the
``Action de recherche concertée'' (ARC) A2, a cooperative research
project initiated and financed by the AUPELF-UREF. The CITI's
machine-aided translation program has since been handed over to the
University of Montreal's RALI laboratory, where we continue
working on the BAF. We have recently made version 1.1 of the corpus
available to the research community. It can be obtained from the RALI,
through its World Wide Web server, at URL
http://www-rali.iro.umontreal.ca.
While such bilingual corpora are now quite common, what distinguishes the BAF from other bitexts is that the alignments were entirely done by hand. As a result, the corpus can be used as a reference to evaluate and compare the performance of various automatic alignment techniques.
The greater part of the corpus is made up of ``institutional'' texts:
debates of the Canadian parliament (Hansards), court transcripts and UN
reports; but we have also included some scientific, technical and
literary documents. In all, the corpus contains approximately
words in each language.
In this paper, we discuss why this corpus was assembled, how it was produced, and what it contains. We also describe some of the computer tools that were developed and used in the process, and which we have also made available.