Structured Learning: Inferring Graphs from Structured and Unstructured Inputs

Abbreviated Title: 
SLG2013
Call for Papers
Submission Deadline: 
15 Apr 2013
Event Dates: 
15 Jun 2013
Location: 
ICML workshop, June 16, 2013 (day between ICML & NAACL)
City: 
Atlanta
State: 
Georgia
Country: 
USA
Contact Email: 
slg-2013-chairs [at] googlegroups [dot] com

Structured Learning: Inferring Graphs from Structured and Unstructured
Inputs (SLG2013)
ICML workshop, June 16, 2013 (day between ICML & NAACL)
https://sites.google.com/site/slgworkshop2013/
Submission Deadline: April 15, 2013

OVERVIEW

Structured learning involves learning and making inferences from inputs
that can be both unstructured (e.g., text) and structured (e.g., graphs
and graph fragments), and making predictions about outputs that are also
structured as graphs. Examples include the construction of knowledge
bases from noisy extraction data, inferring temporal event graphs from
newswire or social media, and inferring influence structure on social
graphs from multiple sources.

One of the challenges of this setting is that it often does not fit into
the classic supervised or unsupervised learning paradigm. In essence,
we have one large (potentially infinite) partially observed input graph,
and we are trying to make inferences about the unknown aspects of this
graph's structure. Often times there is side information available,
which can be used for enrichment, but in order to use this information,
we need to infer mappings for schema and ontologies that describe that
side information, perform alignment and entity resolution, and reason
about the added utility of the additional sources. The topic is
extremely pressing, as many of the modern challenges in extracting
usable knowledge from (big) data fall into this setting. Our focus in
this workshop is on the machine learning and inference methods that are
useful in such settings.

Topics of interest include, but are not limited to:

- Graph-based methods for entity resolutions and word sense disambiguation
- Graph-based representations for ontology learning
- Graph-based strategies for semantic relations identification
- Making use of taxonomies and encoding semantic distances in graphs
- Random walk methods in graphs
- Spectral graph clustering and multi-relational clustering
- Semi-supervised graph-based methods
- Graph summarization

Our goal is to bring together researchers in graphical models,
structured prediction, latent variable relational models and statistical
relational learning in order to (a) exchange experiences applying
various methods to these graph learning domains, (b) share their
successes, and (c) identify common challenges. The outcomes we expect
are (1) a better understanding of the different existing methods across
disparate communities, (2) an identification of common challenges, and
(3) a venue for sharing resources, tools and datasets.

The workshop will consist of a number of invited talks in each of these
areas, a poster session where participants can present their work, and
discussion.

SUBMISSION INFORMATION

We solicit short, poster-length submissions of up to 2 pages. All
accepted submissions will be presented as posters, and a subset of them
may be considered for oral presentation. Submissions reporting work in
progress are acceptable, as we aim for the workshop to offer a venue for
stimulating discussions.

Submissions must be in PDF format, and should be made through Easychair
at https://www.easychair.org/conferences/?conf=slg-2013
Important dates

- Submission deadline: April 15, 2013
- Notification of acceptance: April 30, 2013
- Final versions of accepted submissions due: May 15, 2013
- Workshop date: Sunday, June 16, 2013
(Note: Most of the ICML workshops are taking place AFTER the conference,
while this workshop takes place before the main conference, on the same
day as ICML tutorials, in order to make it easier for participants from
NAACL HLT conference to attend).

ORGANIZING COMMITTEE

- Hal Daume III, University of Maryland
- Evgeniy Gabrilovich, Google
- Lise Getoor, University of Maryland
- Kevin Murphy, Google

PROGRAM COMMITTEE(confirmed to date)

- Jeff Dalton, UMass
- Laura Dietz, UMass
- Thorsten Joachims, Cornell
- Daniel Lowd, University of Oregon
- Mausam, University of Washington
- Jennifer Neville, Purdue University
- Stuart Russell, UC Berkeley
- Ivan Titov, Saarland University
- Daisy Zhe Wang, University of Florida
- Jerry Zhu, University of Wisconsin - Madison

FURTHER INFORMATION

For further information, please contact the workshop organizers at
slg-2013-chairs@googlegroups.com