minor typos in my part

git-svn-id: https://mosesdecoder.svn.sourceforge.net/svnroot/mosesdecoder/trunk@1125 1f5c12ca-751b-0410-a591-d2e778427230

1 files changed, 3 insertions, 3 deletions

diff --git a/report/report.tex b/report/report.tex
index d6147f9da..b4d4d6279 100755
--- a/report/report.tex
+++ b/report/report.tex
@@ -1388,7 +1388,7 @@ A second tool developed during the workshop shows the mapping of individual sour
 
 
 This report describes in detail our experiments on \bidir{} translation
-with the Moses system \parcite{moses} carried out at Johns Hopkins University
+with the Moses system carried out at Johns Hopkins University
 Summer Workshop 2006 in Baltimore. The reader is expected to be
 familiar with factored translation models as implemented in Moses.
 
@@ -1437,7 +1437,7 @@ Total                                  &  1.4 M   &  1.3 M   &  19.0 M  &  21.6
 The texts in CzEng are pre-tokenized and pre-segmented (sentence boundaries identified) and
 automatically sentence-aligned using the hunalign tool
 \parcite{varga:hunalign:2005}. The
-PCEDT data are manually aligned 1-1 by origin, because the Czech
+PCEDT data are manually sentence aligned 1-1 by origin, because the Czech
 version of the text was obtained by translating English text sentence by
 sentence.
 
@@ -1796,7 +1796,7 @@ We also see nearly no gain or even some loss by increasing the corpus size
 from 20k to 860k sentences. (See section \ref{moredata}
 below for more details on various ways of using more data.)
 This observation can be explained by the very specific
-domain of our test set, see section
+domain of our test set.
 
 
 \subsubsection{Scenarios of Factored Translation \tocs{}}