Improving Neural Machine Translation with AMR Semantic Graphs

<table class="table-group" id="tab3"><tr><td><table class="table"><tr><td class="thead-hr" colspan="3"><hr/></td></tr><tr class="thead"><td class="align_left" rowspan="2">Model</td><td class="align_center" colspan="2">BLEU</td></tr><tr class="thead"><td class="align_center">tst2013</td><td class="align_center">tst2015</td></tr><tr><td class="thead-hr" colspan="3"><hr/></td></tr><tr><td class="align_left">LightConvSeq2Seq</td><td class="align_center">27.47</td><td class="align_center">25.09</td></tr><tr><td class="align_left">LightConvSeq2Seq-AMR-F</td><td class="align_center">27.71</td><td class="align_center">25.05</td></tr><tr><td class="align_left">LightConvSeq2Seq-AMR-B</td><td class="align_center">27.84</td><td class="align_center">25.27</td></tr><tr><td class="align_left">LightConvSeq2Seq-AMR</td><td class="align_center"><b>28.46</b></td><td class="align_center"><b>25.67</b></td></tr><tr class="table-tr"><td colspan="3"><hr class="tbody-hr"/></td></tr></table></td></tr><tr class="table-fn"><td><div>The bold values are the highest results when evaluating each model for the “tst2013” and “tst2015” testsets.<br/></div></td></tr></table>

<div>Experimental results on LightConvSeq2Seq.</div>

Mathematical Problems in Engineering

tab3

Table 3

Table 3: Improving Neural Machine Translation with AMR Semantic Graphs 