<div dir="ltr">
<div style="text-decoration-style:initial;text-decoration-color:initial">
<div class="gmail-gE gmail-iv gmail-gt" style="font-size:12.8px;padding:10px 0px 3px;background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial"><b style="font-family:arial,helvetica,sans-serif;font-size:12.8px">When:</b><span style="font-family:arial,helvetica,sans-serif;font-size:12.8px;font-weight:bold"> </span><span style="font-family:arial,helvetica,sans-serif;font-size:12.8px"> Mon</span><span class="gmail-m_-7966131709472399317gmail-aBn" style="font-family:arial,helvetica,sans-serif;font-size:12.8px;border-bottom:1px dashed rgb(204,204,204)"><span class="gmail-m_-7966131709472399317gmail-aQJ" style="">day, June 25th at 10:30 am</span></span><br></div><div class="gmail-" style="background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial"><div id="gmail-:25s" class="gmail-ii gmail-gt" style="direction:ltr;margin:5px 15px 0px 0px;padding-bottom:5px"><div id="gmail-:25t" class="gmail-a3s gmail-aXjCH gmail-m1640024431f2db78" style="overflow:hidden"><div dir="ltr"><div style="font-size:12.8px;text-decoration-style:initial;text-decoration-color:initial"><b><span class="gmail-m_-7966131709472399317gmail-aBn" style="border-bottom:1px dashed rgb(204,204,204)"><span class="gmail-m_-7966131709472399317gmail-aQJ"><font face="arial, helvetica, sans-serif"><br></font></span></span></b></div><div style="font-size:12.8px;text-decoration-style:initial;text-decoration-color:initial"><font face="arial, helvetica, sans-serif"><b>Where: <span> </span></b><span style="font-weight:bold"> </span><span class="gmail-m_-7966131709472399317gmail-aBn" style="border-bottom:1px dashed rgb(204,204,204)"><span class="gmail-m_-7966131709472399317gmail-aQJ">TTIC, 6045 S. Kenwood Avenue, 5th Floor, Room 526</span></span></font></div><div style="font-size:12.8px;text-decoration-style:initial;text-decoration-color:initial"><b><span class="gmail-m_-7966131709472399317gmail-aBn" style="border-bottom:1px dashed rgb(204,204,204)"><span class="gmail-m_-7966131709472399317gmail-aQJ"><font face="arial, helvetica, sans-serif"><br></font></span></span></b></div><div style="text-decoration-style:initial;text-decoration-color:initial"><font face="arial, helvetica, sans-serif" style="font-size:12.8px"><b>Who: </b></font><font face="arial, helvetica, sans-serif"><span style="font-size:12.8px">Greg Durrett, UT Austin</span></font></div></div></div></div></div>
<br></div><div style="font-size:12.8px;text-decoration-style:initial;text-decoration-color:initial"><br></div><div style="font-size:12.8px;text-decoration-style:initial;text-decoration-color:initial"><b>Title:</b> Hand-holding Learning of Latent Structure for Natural Language Processing<br></div><div style="font-size:12.8px;text-decoration-style:initial;text-decoration-color:initial"><span style="background-color:rgb(255,255,255);text-decoration-style:initial;text-decoration-color:initial;float:none;display:inline"><div><br></div><div><b>Abstract:</b> Neural network models are extremely powerful, but can typically fit a training set in many ways, including ways that don't generalize; this is the classic overfitting problems. While techniques like regularization and dropout can help, these still don't explicitly encourage networks to learn the right latent representations for the problem at hand. In this talk, I will discuss some recent work my group has done on "hand-holding" neural network models for NLP: incorporating auxiliary supervision into neural networks or constraining their latent representations to improve generalization. I will primarily focus on two NLP problems involving complex reasoning: document-based reading comprehension and GRE-style math word problem solving. For both problems, our models must learn to perform multiple steps of reasoning based on input text which can be both ambiguous and display high amounts of lexical variation. We incorporate auxiliary supervision derived either from human annotation or a heuristic solver and show that this can improve the performance of a learned model. Finally, I will briefly discuss some current work on choosing the right distribution for latent variables in deep generative models, which imposes a similar kind of prior knowledge about the task's structure.</div></span></div>
<div><br></div><div><br></div><div>Host: <a href="mailto:kgimpel@ttic.edu">Kevin Gimpel</a></div><div><br></div><div><br></div><br clear="all"><div><div class="gmail_signature"><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><font face="arial, helvetica, sans-serif">Mary C. Marre</font><div><font face="arial, helvetica, sans-serif">Administrative Assistant</font></div><div><i><font face="arial, helvetica, sans-serif" color="#3d85c6"><b>Toyota Technological Institute</b></font></i></div><div><i><font face="arial, helvetica, sans-serif" color="#3d85c6">6045 S. Kenwood Avenue</font></i></div><div><i><font face="arial, helvetica, sans-serif" color="#3d85c6">Room 504</font></i></div><div><i><font face="arial, helvetica, sans-serif" color="#3d85c6">Chicago, IL 60637</font></i></div><div><i><font face="arial, helvetica, sans-serif">p:(773) 834-1757</font></i></div><div><i><font face="arial, helvetica, sans-serif">f: (773) 357-6970</font></i></div><div><b><i><a href="mailto:mmarre@ttic.edu" target="_blank"><font face="arial, helvetica, sans-serif">mmarre@ttic.edu</font></a></i></b></div></div></div></div></div></div></div></div></div></div>
</div>