<div>EXTENDED DEADLINE: 15 March, 2009</div><div><br></div><div>Workshop on Integer Linear Programming for NLP at NAACL-HLT 2009</div><div>==========================================================</div><div>NAACL HLT 2009 Workshop on</div>
<div>Integer Linear Programming for Natural Language Processing</div><div><br></div><div>June 4, 2009, Boulder, Colorado, USA</div><div><a href="http://www-tsujii.is.s.u-tokyo.ac.jp/ilpnlp/">http://www-tsujii.is.s.u-tokyo.ac.jp/ilpnlp/</a></div>
<div><br></div><div>Call for Papers</div><div>(Submission deadline: March 15, 2009)</div><div>==========================================================</div><div>Integer Linear Programming (ILP) has recently attracted much attention</div>
<div>within the NLP community. Formulating problems using ILP has several</div><div>advantages. It allows us to focus on the modelling of problems,</div><div>rather than engineering new search algorithms; provides the</div>
<div>opportunity to incorporate generic global constraints; and guarantees</div><div>exact inference. This and the availability of off-the-shelf solvers</div><div>has lead to a large variety of natural language processing tasks being</div>
<div>formulated in the ILP framework, including semantic role labelling,</div><div>syntactic parsing, summarisation and joint information extraction.</div><div><br></div><div>The use of ILP brings many benefits and opportunities but there are</div>
<div>still challenges for the community; these include: formulations of new</div><div>applications, dealing with large-scale problems and understanding the</div><div>interaction between learning and inference at training and decision</div>
<div>time. The purpose of this workshop is to bring together researchers</div><div>interested in exploiting ILP for NLP applications and tackling the</div><div>issues involved. We are interested in a broad range of topics</div>
<div>including, but not limited to:</div><div><br></div><div>- Novel ILP formulations of NLP tasks. This includes: the</div><div>introduction of ILP formulations of tasks yet to be tackled within the</div><div>framework; and novel formulations, such as equivalent LP relaxations,</div>
<div>that are more efficient to process than previous formulations.</div><div><br></div><div>- Learning and Inference. This includes issues relating to:</div><div>decoupling of learning (e.g., learning through local classifiers) and</div>
<div>inference, learning with exact (e.g., ILP) or approximate inference,</div><div>learning of constraints, learning weights for soft constraints, and</div><div>the impact of ignoring various constraints during learning.</div>
<div><br></div><div>- The utility of global hard and soft constraints in NLP. Sometimes</div><div>constraints do not increase accuracy (and can even decrease it), when</div><div>and why do global constraints become useful? For example, do global</div>
<div>constraints become more important if we have less data?</div><div><br></div><div>- Formulating and solving large NLP problems. Applying ILP to hard</div><div>problems (such as parsing, machine translation and solving several NLP</div>
<div>tasks at once) often results in very large formulations which can be</div><div>impossible to solve directly by the ILP engine. This may require</div><div>exploring different ILP solving methods (such as, approximate ILP</div>
<div>solvers/methods) or cutting plane and pricing techniques.</div><div><br></div><div>- Alternative declarative approaches. A variety of other modeling</div><div>frameworks exist, of which ILP is just one instance. Using other</div>
<div>approaches, such as weighted MAX-SAT, Constraint Satisfaction Problems</div><div>(CSP) or Markov Networks, could be more suitable than ILP in some</div><div>cases. It can also be helpful to model a problem in one framework</div>
<div>(e.g., Markov Networks) and solve them with another (e.g., ILP) by</div><div>using general mappings between representations.</div><div><br></div><div>- First Order Modelling Languages. ILP, and other essentially</div>
<div>propositional languages, require the creation of wrapper code to</div><div>generate an ILP formulation for each problem instance. First (Higher)</div><div>Order languages, such as Learning Based Java and Markov Logic, reduce</div>
<div>this overhead and can also aid the solver to be more efficient.</div><div>Moreover, with such languages the automatic exploration of the model</div><div>space is easier.</div><div><br></div><div>SUBMISSION INFORMATION</div>
<div><br></div><div>We encourage submissions addressing the above questions and topics or</div><div>other relevant issues. Authors are invited to submit a full paper of</div><div>up to 8 pages (with up to 1 additional page for references), or an</div>
<div>abstract of up to 2 pages. Appropriate topics for abstracts include</div><div>preliminary results, application notes, descriptions of work in</div><div>progress, etc. Previously published papers cannot be accepted.</div>
<div><br></div><div>The submissions will be reviewed by the program committee. Note that</div><div>reviewing will be blind and hence no author information should be</div><div>included in the papers. Self-references that reveal the author's</div>
<div>identity, e.g., "We previously showed (Smith, 1991) …", should be</div><div>avoided. Instead, use citations such as "Smith previously showed</div><div>(Smith, 1991) …".</div><div><br></div><div>Papers will be accepted on or before 6 March 2009 in PDF format via</div>
<div>the START system at <a href="https://www.softconf.com/naacl-hlt09/ILPNLP2009/">https://www.softconf.com/naacl-hlt09/ILPNLP2009/</a>.</div><div>Submissions should follow the NAACL HLT 2009 formatting requirements</div>
<div>for full papers , found at</div><div><a href="http://clear.colorado.edu/NAACLHLT2009/stylefiles.html">http://clear.colorado.edu/NAACLHLT2009/stylefiles.html</a>.</div><div><br></div><div>IMPORTANT DATES:</div><div>March 15, 2009: Submission deadline</div>
<div>March 30, 2009: Notification of acceptance</div><div>April 12, 2009: Camera-ready copies due</div><div>June 4, 2009: Workshop held in conjunction with NAACL HLT</div><div><br></div><div>INVITED SPEAKER: Dan Roth (University of Illinois at Urbana-Champaign)</div>
<div><br></div><div>PROGRAM COMMITTEE:</div><div>- Dan Roth (University of Illinois at Urbana-Champaign)</div><div>- Mirella Lapata (University of Edinburgh)</div><div>- Scott Yih (Microsoft Research)</div><div>- Nick Rizzolo (University of Illinois at Urbana-Champaign)</div>
<div>- Ming-Wei Chang (University of Illinois at Urbana-Champaign)</div><div>- Ivan Meza-Ruiz (University of Edinburgh)</div><div>- Ryan McDonald (Google Research)</div><div>- Jenny Rose Finkel (Stanford University)</div>
<div>- Pascal Denis (INRIA Paris-Rocquencourt)</div><div>- Manfred Klenner (University of Zurich)</div><div>- Hal Daume III (University of Utah)</div><div>- Daniel Marcu (University of Southern California)</div><div>- Kevin Knight (University of Southern California)</div>
<div>- Katja Filippova (EML Research)</div><div>- Mark Dras (Macquarie University)</div><div>- Hiroya Takamura (Tokyo Institute of Technology)</div><div><br></div><div>ORGANIZERS AND CONTACT:</div><div>- James Clarke (University of Illinois at Urbana-Champaign)</div>
<div>- Sebastian Riedel (University of Tokyo)</div><div><br></div><div>Email: <a href="mailto:ilpnlp2009@gmail.com">ilpnlp2009@gmail.com</a></div><div>Website: <a href="http://www-tsujii.is.s.u-tokyo.ac.jp/ilpnlp/">http://www-tsujii.is.s.u-tokyo.ac.jp/ilpnlp/</a></div>