tech,4-1-P05-1058,ak models </term> . This paper proposes an <term> alignment adaptation approach </term> to improve <term> domain-specific (
tech,9-1-P05-1058,ak adaptation approach </term> to improve <term> domain-specific ( in-domain ) word alignment </term> . The basic idea of <term> alignment
tech,4-2-P05-1058,ak alignment </term> . The basic idea of <term> alignment adaptation </term> is to use <term> out-of-domain corpus
lr,9-2-P05-1058,ak alignment adaptation </term> is to use <term> out-of-domain corpus </term> to improve <term> in-domain word alignment
other,13-2-P05-1058,ak out-of-domain corpus </term> to improve <term> in-domain word alignment results </term> . In this paper , we first train
model,8-3-P05-1058,ak In this paper , we first train two <term> statistical word alignment models </term> with the <term> large-scale out-of-domain
lr,14-3-P05-1058,ak word alignment models </term> with the <term> large-scale out-of-domain corpus </term> and the <term> small-scale in-domain
lr,19-3-P05-1058,ak out-of-domain corpus </term> and the <term> small-scale in-domain corpus </term> respectively , and then interpolate
model,29-3-P05-1058,ak respectively , and then interpolate these two <term> models </term> to improve the <term> domain-specific
tech,33-3-P05-1058,ak two <term> models </term> to improve the <term> domain-specific word alignment </term> . Experimental results show that
tech,7-4-P05-1058,ak results show that our approach improves <term> domain-specific word alignment </term> in terms of both <term> precision </term>
measure(ment),14-4-P05-1058,ak word alignment </term> in terms of both <term> precision </term> and <term> recall </term> , achieving
measure(ment),16-4-P05-1058,ak terms of both <term> precision </term> and <term> recall </term> , achieving a <term> relative error
measure(ment),20-4-P05-1058,ak and <term> recall </term> , achieving a <term> relative error rate reduction </term> of 6.56 % as compared with the state-of-the-art
skrij podrobnosti