<div dir="ltr">Dear all,<div><br></div><div><br></div><div>I have a question about finding the best phrase alignments. The alignments are used by MOSES during the decoding phrase. </div><div><br></div><div> I have a pair parallel sentences ( a source / a target). I need the best phrase alignment between the source and target sentences. The best phrase alignment is a alignment that MOSES is used to translate the source sentence to the target sentence.<br>
</div><div><br></div><div>Let me use an example to explain what i want.<br></div><div>Example:</div><div style> I have a pair sentence:</div><div>Source : I go to the home </div><div>Target : man be khaneh miravam (in farsi) </div>
<div><br></div><div style>I need the following alignment:</div><div><br></div><div>The Best alignment : [I-->man] [to the --> beh] [home-->khaneh] [ go --> miravam]<br clear="all"><div style>the result includes two sort of information</div>
<div style><br></div><div style>1- the best segments</div><div style>2- the best alignment</div><div style><br></div><div style><div>We can use MOSES for extracting the alignments when using training sentences as input sentence of the MOses's decoder. But there is problem. The output of the decoder is not exactly same as the target sentence.</div>
<div><br></div></div><div style>I know that the giza++ is used for word alignments. I need a solution for phrase alignments. </div><div style>Tnx </div>-- <br> S.Farzi, Ph.D. Student <br> Natural Language Processing Lab,<br>
School of Electrical and Computer Eng.,<br> Tehran University<br> Tel: +9821-6111-9719<br><br>
</div></div>