<?xml version="1.0" encoding="UTF-8" ?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
	<title>Posts - SFU Natural Language Laboratory</title>
	<link>http://natlang.cs.sfu.ca/</link>
	<description>
		Posts - for the Natural Laboratory Laboratory at Simon Fraser University, School of Computing Science.
	</description>
	<lastBuildDate>Mon, 17 Jun 2013 21:00:07 -0700</lastBuildDate>

	
<item>
	<title>SFU Natlang at NAACL HLT 2013 in Atlanta, GA</title>
	<description>
	&lt;p&gt;Anoop Sarkar and Baskaran Sankaran will be attending &lt;span class=&quot;caps&quot;&gt;NAACL&lt;/span&gt; &lt;span class=&quot;caps&quot;&gt;HLT&lt;/span&gt; 2013 in Atlanta, GA from Jun 9-13, 2013.&lt;/p&gt;
&lt;p&gt;Baskaran will be presenting his paper on &lt;em&gt;Multi-metric optimization using Ensemble Tuning&lt;/em&gt; on Wednesday, Jun 12th, 2013 in the W1a Session (Machine Translation) at 11:05am.&lt;/p&gt;
&lt;p&gt;Hope to meet you there!&lt;/p&gt;
	</description>
	<guid>http://natlang.cs.sfu.ca/posts/2013-06-08-SFU-Natlang-at-NAACL-HLT-2013-in-Atlanta%2C-GA</guid>
	<link>http://natlang.cs.sfu.ca/posts/2013-06-08-SFU-Natlang-at-NAACL-HLT-2013-in-Atlanta%2C-GA</link>
	<pubDate>Sat, 08 Jun 2013 00:00:00 -0700</pubDate>
</item>

<item>
	<title>Multi-Metric Optimization Using Ensemble Tuning</title>
	<description>
	&lt;p&gt;In the lab meeting on Wednesday, 29 May, Baskaran Sankaran will give a talk about his NAACL 2013 paper, titled, &amp;#8220;Multi-Metric Optimization Using Ensemble Tuning&amp;#8221;. A brief description is below :&lt;/p&gt;

&lt;p&gt;&amp;#8220;This paper examines tuning for statistical machine translation (SMT) with respect to multiple evaluation metrics. We propose several novel methods for tuning towards multiple objectives, including some based on ensemble decoding methods. Pareto-optimality is a natural way to think about multi-metric optimization (MMO) and our methods can effectively combine several Pareto-optimal solutions, obviating the need to choose one. Our best performing ensemble tuning method is a new algorithm for multi-metric optimization that searches for Pareto-optimal ensemble models. We study the effectiveness of our methods through experiments on multiple as well as single reference(s) datasets. Our experiments show simultaneous gains across several metrics (BLEU, RIBES), without any significant reduction in other metrics. This contrasts the traditional tuning where gains are usually limited to a single metric. Our human evaluation results confirm that in order to produce better MT output, optimizing multiple metrics is better than optimizing only one&amp;#8221;&lt;/p&gt;

&lt;p&gt;The lab meeting will start at 1230 hours at TASC1 9408.&lt;/p&gt;
	</description>
	<guid>http://natlang.cs.sfu.ca/posts/2013-05-29-Baskaran-Talk</guid>
	<link>http://natlang.cs.sfu.ca/posts/2013-05-29-Baskaran-Talk</link>
	<pubDate>Wed, 29 May 2013 00:00:00 -0700</pubDate>
</item>

<item>
	<title>Press release about our lab research</title>
	<description>
	&lt;p&gt;On May 22, 2013 &lt;span class=&quot;caps&quot;&gt;SFU&lt;/span&gt; had &lt;a href=&quot;http://www.sfu.ca/pamr/media-releases/2013/advancing-research-on-lifes-origins-language-translation.html&quot;&gt;a press release about our lab research&lt;/a&gt; and the Natural Sciences and Engineering Research Council (&lt;span class=&quot;caps&quot;&gt;NSERC&lt;/span&gt;) Discovery Accelerator Supplement (&lt;span class=&quot;caps&quot;&gt;DAS&lt;/span&gt;) award to Anoop Sarkar.&lt;/p&gt;
	</description>
	<guid>http://natlang.cs.sfu.ca/posts/2013-05-23-Press-release-about-our-lab-research</guid>
	<link>http://natlang.cs.sfu.ca/posts/2013-05-23-Press-release-about-our-lab-research</link>
	<pubDate>Thu, 23 May 2013 00:00:00 -0700</pubDate>
</item>

<item>
	<title>Lab lunch</title>
	<description>
	&lt;p&gt;We are doing lab lunch at &lt;a href='http://bandidastaqueria.com/'&gt;Bandidas Taqueria&lt;/a&gt; on Wednesday May 22 at 12:00 noon, because Max is leaving. This will replace the usual lab meeting. Bandidas is located at &lt;a href='http://maps.google.com/maps?f=q&amp;amp;hl=en&amp;amp;geocode=&amp;amp;q=2781+Commercial+Drive,+Vancouver,+bc&amp;amp;sll=37.0625,-95.677068&amp;amp;sspn=37.325633,57.216797&amp;amp;ie=UTF8&amp;amp;ll=49.268365,-123.065672&amp;amp;spn=0.00752,0.013969&amp;amp;z=14&amp;amp;g=2781+Commercial+Drive,+Vancouver,+bc&amp;amp;iwloc=addr&amp;amp;source=embed'&gt;2781 Commercial Drive, Vancouver&lt;/a&gt;.&lt;/p&gt;
	</description>
	<guid>http://natlang.cs.sfu.ca/posts/2013-05-21-Lunch-because-Max-is-leaving</guid>
	<link>http://natlang.cs.sfu.ca/posts/2013-05-21-Lunch-because-Max-is-leaving</link>
	<pubDate>Tue, 21 May 2013 00:00:00 -0700</pubDate>
</item>

<item>
	<title>Lab meeting time</title>
	<description>
	&lt;p&gt;Please note that for this term, the lab meetings will be held from 1230 hours every Wednesday. The venue for the lab meeting remains the same, TASC1 9408.&lt;/p&gt;
	</description>
	<guid>http://natlang.cs.sfu.ca/posts/2013-05-13-Lab-meeting</guid>
	<link>http://natlang.cs.sfu.ca/posts/2013-05-13-Lab-meeting</link>
	<pubDate>Mon, 13 May 2013 00:00:00 -0700</pubDate>
</item>

<item>
	<title>Emergence of Gricean Maxims from Multi-Agent Decision Theory</title>
	<description>
	&lt;p&gt;In the lab meeting on Wednesday, May 8, Ann Clifton is going to discuss the paper &lt;a href='http://nlp.stanford.edu/pubs/cards-naacl2013.pdf'&gt;Emergence of Gricean Maxims from Multi-Agent Decision Theory&lt;/a&gt;. A short description is as follows:&lt;/p&gt;

&lt;p&gt;&amp;#8221; This paper trains multiple agents to perform a cooperative task using linguistic information and pragmatic reasoning through a decentralized partially observable markov decision process model. The authors argue that the models learned exhibit the properties of the &amp;#8220;cooperative principle&amp;#8221;, a pragmatic linguistic theory of how natural language conversation is conducted. &amp;#8221;&lt;/p&gt;

&lt;p&gt;The lab meeting will start at 1100 hours at TASC1 9408.&lt;/p&gt;
	</description>
	<guid>http://natlang.cs.sfu.ca/posts/2013-05-08-Ann</guid>
	<link>http://natlang.cs.sfu.ca/posts/2013-05-08-Ann</link>
	<pubDate>Wed, 08 May 2013 00:00:00 -0700</pubDate>
</item>

<item>
	<title>Marginalized Denoising Autoencoders for Domain Adaptation</title>
	<description>
	&lt;p&gt;In the lab meeting on Wednesday, April 24, 2013, Majid Razmara will give a talk on the paper &lt;a href='http://arxiv.org/abs/1206.4683'&gt;Marginalized Denoising Autoencoders for Domain Adaptation&lt;/a&gt;. The meeting will be at TASC1 9408 and start at 1100 hours.&lt;/p&gt;
	</description>
	<guid>http://natlang.cs.sfu.ca/posts/2013-04-24-Majid-marginalized</guid>
	<link>http://natlang.cs.sfu.ca/posts/2013-04-24-Majid-marginalized</link>
	<pubDate>Wed, 24 Apr 2013 00:00:00 -0700</pubDate>
</item>

<item>
	<title>Similarity Estimation Techniques from Rounding Algorithms</title>
	<description>
	&lt;p&gt;In the lab meeting on 10 April, 2013, Bruce Krayenhoff will give a talk. The details are below :&lt;/p&gt;

&lt;p&gt;We will discuss the first-half of the paper &lt;a href='http://www.cs.princeton.edu/~moses/papers/similar.ps'&gt;Similarity Estimation Techniques from Rounding Algorithms.&lt;/a&gt; &amp;#8216;Locality sensitive hashing schemes&amp;#8217; enable us to generate hashing functions h such that the probability that h(x)=h(y) is equal to some measure of the similarity between x and y. I will cover what locality sensitive hashing schemes are and what properties they have. In addition, we will consider one such hashing scheme for vectors.&lt;/p&gt;

&lt;p&gt;Finally we will consider the definition of the EMD (Earth Mover Distance), and how it can be interpreted. The paper discusses how to use hashing functions to approximate the EMD, however I will be leaving this to a future lab meeting.&lt;/p&gt;

&lt;p&gt;The lab meeting will be at TASC1 9408 and start at 1100 hours.&lt;/p&gt;
	</description>
	<guid>http://natlang.cs.sfu.ca/posts/2013-04-10-Bruce-talk</guid>
	<link>http://natlang.cs.sfu.ca/posts/2013-04-10-Bruce-talk</link>
	<pubDate>Wed, 10 Apr 2013 00:00:00 -0700</pubDate>
</item>

<item>
	<title>Evaluating question generation output</title>
	<description>
	&lt;p&gt;David Lindberg will give a talk in the lab meeting on Wednesday, April 3 at TASC1 9408. A short description is below :&lt;/p&gt;

&lt;p&gt;&lt;em&gt;I will begin with a quick review of my question generation(QG) research, describing the context, scope and goals. After discussing previous efforts in evaluating the output of question generation systems, I will conclude with a discussion of the results and implications of an evaluation of my QG system&amp;#8217;s output&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;The talk will begin at 1100 hours.&lt;/p&gt;
	</description>
	<guid>http://natlang.cs.sfu.ca/posts/2013-04-03-David-Lindberg</guid>
	<link>http://natlang.cs.sfu.ca/posts/2013-04-03-David-Lindberg</link>
	<pubDate>Wed, 03 Apr 2013 00:00:00 -0700</pubDate>
</item>

<item>
	<title>Hope and Fear in SMT</title>
	<description>
	&lt;p&gt;We will discuss the David Chiang paper titled, &lt;a href='http://www.isi.edu/~chiang/papers/chiang-jmlr12-corrected.pdf'&gt;Hope and Fear for Discriminative Training of Statistical Translation Models&lt;/a&gt;, in the lab meeting on 26 March. It will start from 1100 hours at TASC1 9408.&lt;/p&gt;
	</description>
	<guid>http://natlang.cs.sfu.ca/posts/2013-03-27-David-Chiang</guid>
	<link>http://natlang.cs.sfu.ca/posts/2013-03-27-David-Chiang</link>
	<pubDate>Wed, 27 Mar 2013 00:00:00 -0700</pubDate>
</item>


</channel>
</rss>
