From dyogatama at cs.cmu.edu  Mon Mar 31 17:39:38 2014
From: dyogatama at cs.cmu.edu (Dani Yogatama)
Date: Mon, 31 Mar 2014 17:39:38 -0400
Subject: [CL+NLP Lunch] CL+NLP Lunch, Dan Garrette, April 10 @noon
Message-ID: <CAJPxrhwYNMh_mORpL0x_B5FVp-SOVUHkxf7nZmsKsCTWO_aRfA@mail.gmail.com>

*CL+NLP Lunch *(*http://www.cs.cmu.edu/~nlp-lunch/
<http://www.cs.cmu.edu/~nlp-lunch/>*)
*Speaker*: Dan Garrette, The University of Texas at Austin
*Date*: Thursday, April 10, 2013
*Time*: 12:00 noon
*Venue*: GHC 6501

*Title*: Unsupervised learning of non+concatenative morphology

*Abstract*:
Grammar learning is a well-studied problem in NLP, but the task is
particularly difficult for low-resource languages.  In this talk, I
will discuss our current work in learning combinatory categorial
grammars from various forms of weak supervision.  First, I will show
how we can learn good sequence-based CCG supertaggers by encoding
universal, inherent properties of the CCG formalism as priors over
both the appearance of supertags and the transitions between
supertags.  These universal priors can, in turn, be combined with
corpus-specific knowledge derived from available (partial) tag
dictionaries and unannotated text to further improve tagging
performance.  Then, I will discuss our current efforts to extend these
principles to tree grammars to learn CCG parsers.  Finally, I will
discuss how simple annotations --- particularly those given in the
Graph Fragment Language developed at CMU --- may be used to help
learn parsers under extremely tight annotation budgets.

This work is in collaboration with Jason Baldridge, Chris Dyer, and Noah
Smith.

*Biography*:
Dan is a a Computer Science Ph.D. student at The University of Texas
at Austin. His research focuses on Natural Language Processing and
Machine Learning.  He was a Best Talk Award nominee at NAACL this
past year. He thinks slides should have less text and more animations.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/nlp-lunch/attachments/20140331/8a89d11b/attachment.html>

From dyogatama at cs.cmu.edu  Thu Apr 10 00:03:46 2014
From: dyogatama at cs.cmu.edu (Dani Yogatama)
Date: Thu, 10 Apr 2014 00:03:46 -0400
Subject: [CL+NLP Lunch] CL+NLP Lunch, Dan Garrette, April 10 @noon
In-Reply-To: <CAJPxrhwYNMh_mORpL0x_B5FVp-SOVUHkxf7nZmsKsCTWO_aRfA@mail.gmail.com>
References: <CAJPxrhwYNMh_mORpL0x_B5FVp-SOVUHkxf7nZmsKsCTWO_aRfA@mail.gmail.com>
Message-ID: <CAJPxrhz8bhfHufd=hzuiS1=q6H5mEOhBo-PHPq1Tuc8+5ovMEQ@mail.gmail.com>

reminder tomorrow at noon

On Monday, March 31, 2014, Dani Yogatama <dyogatama at cs.cmu.edu> wrote:

> *CL+NLP Lunch *(*http://www.cs.cmu.edu/~nlp-lunch/
> <http://www.cs.cmu.edu/~nlp-lunch/>*)
> *Speaker*: Dan Garrette, The University of Texas at Austin
> *Date*: Thursday, April 10, 2013
> *Time*: 12:00 noon
> *Venue*: GHC 6501
>
> *Title*:  Learning Combinatory Categorial  Grammars from Wrak Supervision
>
> *Abstract*:
> Grammar learning is a well-studied problem in NLP, but the task is
> particularly difficult for low-resource languages.  In this talk, I
> will discuss our current work in learning combinatory categorial
> grammars from various forms of weak supervision.  First, I will show
> how we can learn good sequence-based CCG supertaggers by encoding
> universal, inherent properties of the CCG formalism as priors over
> both the appearance of supertags and the transitions between
> supertags.  These universal priors can, in turn, be combined with
> corpus-specific knowledge derived from available (partial) tag
> dictionaries and unannotated text to further improve tagging
> performance.  Then, I will discuss our current efforts to extend these
> principles to tree grammars to learn CCG parsers.  Finally, I will
> discuss how simple annotations --- particularly those given in the
> Graph Fragment Language developed at CMU --- may be used to help
> learn parsers under extremely tight annotation budgets.
>
> This work is in collaboration with Jason Baldridge, Chris Dyer, and Noah
> Smith.
>
> *Biography*:
> Dan is a a Computer Science Ph.D. student at The University of Texas
> at Austin. His research focuses on Natural Language Processing and
> Machine Learning.  He was a Best Talk Award nominee at NAACL this
> past year. He thinks slides should have less text and more animations.
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/nlp-lunch/attachments/20140410/5b3e47cf/attachment.html>

From cdyer at cs.cmu.edu  Thu May  8 14:28:14 2014
From: cdyer at cs.cmu.edu (Chris Dyer)
Date: Thu, 8 May 2014 14:28:14 -0400
Subject: [CL+NLP Lunch] CL-NLP-MT Lunch, Stefan Riezler, May 13 @ noon
Message-ID: <CAEHEvxPBfrnhSSH_tGZzXMOwBMPxeNFcuY6n+LeKKxgXPgLkDQ@mail.gmail.com>

*Speaker*: Stefan Riezler, Heidelberg University
(http://www.cl.uni-heidelberg.de/~riezler/)
*Date*: Tuesday, May 13, 2014
*Time*: 12:00 noon
*Venue*: GHC 6501

*Title: *Theoretical and Practical Grounding in Empirical Computational
Linguistics

*Abstract*:
Philosophy of science has pointed out a circularity problem in empirical
sciences that arises if all known measuring procedures for a quantity of a
theory presuppose the validity of this theory. We discuss how this problem
relates to empirical computational linguistics, and define a criterion of
T-non-theoretical grounding as guidance to avoid such circularities. We
exemplify how this criterion can be met by crowdsourcing, task-related data
annotation, or data in the wild. In particular, we illustrate the benefits
of grounded learning in the area of statistical machine translation, e.g.,
by grounding machine translation in semantic parsing and in
cross-lingual information retrieval.

*Bio*: Prof. Stefan Riezler has been appointed full professor and head of
the chair of Linguistic Informatics at Heidelberg University in 2010, after
spending a decade in the world?s most renowned industry research labs
(Xerox PARC, Google). He received his PhD in Computational Linguistics from
the University of T?bingen in 1998, and then conducted post-doctoral work
at Brown University in 1999. Prof. Riezler's research focus is on machine
learning and statistics applied to natural language processing problems,
especially for the application areas of natural-language based web search
and statistical machine translation.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/nlp-lunch/attachments/20140508/89969174/attachment.html>

From cdyer at cs.cmu.edu  Mon May 12 17:41:32 2014
From: cdyer at cs.cmu.edu (Chris Dyer)
Date: Mon, 12 May 2014 17:41:32 -0400
Subject: [CL+NLP Lunch] CL-NLP-MT Lunch, Stefan Riezler, May 13 @ noon
In-Reply-To: <CAEHEvxPBfrnhSSH_tGZzXMOwBMPxeNFcuY6n+LeKKxgXPgLkDQ@mail.gmail.com>
References: <CAEHEvxPBfrnhSSH_tGZzXMOwBMPxeNFcuY6n+LeKKxgXPgLkDQ@mail.gmail.com>
Message-ID: <CAEHEvxPwLzHp=AVF91K4W7HSRUR8BXTWSDaR0aSNj0atm5ptaA@mail.gmail.com>

Reminder everyone: this is Tuesday at 12 in 6501. Come welcome our new
visitor, Stefan!


On Thu, May 8, 2014 at 2:28 PM, Chris Dyer <cdyer at cs.cmu.edu> wrote:

> *Speaker*: Stefan Riezler, Heidelberg University
> (http://www.cl.uni-heidelberg.de/~riezler/)
> *Date*: Tuesday, May 13, 2014
> *Time*: 12:00 noon
> *Venue*: GHC 6501
>
> *Title: *Theoretical and Practical Grounding in Empirical Computational
> Linguistics
>
> *Abstract*:
> Philosophy of science has pointed out a circularity problem in empirical
> sciences that arises if all known measuring procedures for a quantity of a
> theory presuppose the validity of this theory. We discuss how this problem
> relates to empirical computational linguistics, and define a criterion of
> T-non-theoretical grounding as guidance to avoid such circularities. We
> exemplify how this criterion can be met by crowdsourcing, task-related data
> annotation, or data in the wild. In particular, we illustrate the benefits
> of grounded learning in the area of statistical machine translation, e.g.,
> by grounding machine translation in semantic parsing and in
> cross-lingual information retrieval.
>
> *Bio*: Prof. Stefan Riezler has been appointed full professor and head of
> the chair of Linguistic Informatics at Heidelberg University in 2010, after
> spending a decade in the world?s most renowned industry research labs
> (Xerox PARC, Google). He received his PhD in Computational Linguistics from
> the University of T?bingen in 1998, and then conducted post-doctoral work
> at Brown University in 1999. Prof. Riezler's research focus is on machine
> learning and statistics applied to natural language processing problems,
> especially for the application areas of natural-language based web search
> and statistical machine translation.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/nlp-lunch/attachments/20140512/a2b0c6e3/attachment.html>

From nathan at cmu.edu  Wed Jun 18 16:38:14 2014
From: nathan at cmu.edu (Nathan Schneider)
Date: Wed, 18 Jun 2014 16:38:14 -0400
Subject: [CL+NLP Lunch] Tim Baldwin talk: Friday at noon
Message-ID: <CADQLQrXrS76GnLmOQGspv-uYg_KJTmB7PvK5V5vdLyjs1c-AWw@mail.gmail.com>

All,

Join us Friday at noon for a talk by a very special visitor. Lunch will be
provided.

Cheers,
Nathan


NLP Lunch <http://www.cs.cmu.edu/~nlp-lunch/>
Friday, June 20 at noon
GHC 8102

Tim Baldwin, University of Melbourne

*Text Analysis of Social Media Beyond Twitterdome*

Abstract:

Text processing of social media data in NLP has largely centred around
Twitter, with other social media types getting relatively little attention.
In this talk, I will first present an empirical comparison of text sourced
from a range of social media sites (microblogs, comments, user forums,
blogs and Wikipedia), focusing on the relative "noisiness" and diversity of
the linguistic content. I will then discuss the application of discourse
parsing to user forum threads, and incorporation of the results to improve
search quality.

Bio:

Tim Baldwin is a Professor in the Department of Computing and Information
Systems, The University of Melbourne, an Australian Research Council Future
Fellow, and a contributed research staff member of NICTA Victoria. He has
previously held visiting positions at the University of Washington,
University of Tokyo, Saarland University, NTT Communication Science
Laboratories and National Institute of Informatics. His research interests
include text mining of social media, computational lexical semantics,
information extraction and web mining, with a particular interest in the
interface between computational and theoretical linguistics. Current
projects include web user forum mining, text mining of Twitter, and machine
translation evaluation. He is currently Secretary of the Australasian
Language Technology Association and a member of the Executive Committee of
the Asian Federation of Natural Language Processing, and was recently PC
Chair of EMNLP 2013 and *SEM 2013.

Tim completed a BSc(CS/Maths) and BA(Linguistics/Japanese) at The
University of Melbourne in 1995, and an MEng(CS) and PhD(CS) at the Tokyo
Institute of Technology in 1998 and 2001, respectively. Prior to joining
The University of Melbourne in 2004, he was a Senior Research Engineer at
the Center for the Study of Language and Information, Stanford University
(2001-2004).
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/nlp-lunch/attachments/20140618/869d2805/attachment.html>

From dcard at andrew.cmu.edu  Mon Jul 14 16:24:09 2014
From: dcard at andrew.cmu.edu (Dallas Card)
Date: Mon, 14 Jul 2014 16:24:09 -0400
Subject: [CL+NLP Lunch] CL+NLP Lunch, Walid Magdy, July 18 @ 12:30pm
Message-ID: <70a207f3cf35a8de4b155a0ccc257eb5.squirrel@webmail.andrew.cmu.edu>

CL+NLP lunch <http://www.cs.cmu.edu/~nlp-lunch/>
Friday, July 18 at 12:30pm
GHC 8012

Walid Magdy, Qatar Computing Research Institute

*Social Media as a Source of Unbiased News*

Abstract:
News media are usually biased toward some political views. Also, the
coverage of news is limited to news reported by news agencies. Social
media is currently a hub for users to report and discuss news. This
includes news reported or missed by news media. Developing a system that
can generate news reports from social media can give a global unbiased
view on what is hot in a given region. In this talk, the challenges for
tracking topics related to news are discussed. An automatically adapting
information filtering approach is presented that allows tracking broad and
dynamic topics in social media. TweetMogaz, a news portal platform that
generated news from Twitter, is demoed. TweetMogaz reports in real-time
what is happening in hot regions in the middle east, such as Syria and
Egypt, in the form of comprehensive reports that include top tweets,
images, videos, and news article shared by users on Twitter.

Bio:
Walid Magdy is a scientist at the Qatar Computing Research Institute
(QCRI) in Doha, Qatar. Walid's main fields of expertise are information
retrieval and natural language processing. He received his PhD in 2012
from the School of Computing in Dublin City University (DCU), Dublin,
Ireland. Earlier, he received his BSc. and Msc. from the Faculty of
Engineering, Cairo University, Egypt in 2005 and 2008 respectively. He
worked before as research engineer in Microsoft and IBM. Walid has a list
of publications in top tier conferences and journals, such as SIGIR, CIKM,
EMNLP, ICWSM, ECIR, IRJ, and TOIS. He also has a set of filed and issued
patents during his work in IBM, Microsoft, DCU, and QCRI.




From dcard at andrew.cmu.edu  Fri Jul 18 09:52:33 2014
From: dcard at andrew.cmu.edu (Dallas Card)
Date: Fri, 18 Jul 2014 09:52:33 -0400
Subject: [CL+NLP Lunch] Reminder: Today at 12:30pm: CL+NLP Lunch, Walid Magdy,
	*GHC 8102*
Message-ID: <deac3daa473b22c4eb1594c19cc26d65.squirrel@webmail.andrew.cmu.edu>

Reminder: Today at 12:30pm!

Please join us this Friday for a talk by Walid Magdy from the Qatar
Computing Research Institute. Lunch will be provided!


CL+NLP lunch <http://www.cs.cmu.edu/~nlp-lunch/>
Friday, July 18 at 12:30pm
GHC 8102

Walid Magdy, Qatar Computing Research Institute

*Social Media as a Source of Unbiased News*

Abstract:
News media are usually biased toward some political views. Also, the
coverage of news is limited to news reported by news agencies. Social
media is currently a hub for users to report and discuss news. This
includes news reported or missed by news media. Developing a system that
can generate news reports from social media can give a global unbiased
view on what is hot in a given region. In this talk, the challenges for
tracking topics related to news are discussed. An automatically adapting
information filtering approach is presented that allows tracking broad and
dynamic topics in social media. TweetMogaz, a news portal platform that
generated news from Twitter, is demoed. TweetMogaz reports in real-time
what is happening in hot regions in the middle east, such as Syria and
Egypt, in the form of comprehensive reports that include top tweets,
images, videos, and news article shared by users on Twitter.

Bio:
Walid Magdy is a scientist at the Qatar Computing Research Institute
(QCRI) in Doha, Qatar. Walid's main fields of expertise are information
retrieval and natural language processing. He received his PhD in 2012
from the School of Computing in Dublin City University (DCU), Dublin,
Ireland. Earlier, he received his BSc. and Msc. from the Faculty of
Engineering, Cairo University, Egypt in 2005 and 2008 respectively. He
worked before as research engineer in Microsoft and IBM. Walid has a list
of publications in top tier conferences and journals, such as SIGIR, CIKM,
EMNLP, ICWSM, ECIR, IRJ, and TOIS. He also has a set of filed and issued
patents during his work in IBM, Microsoft, DCU, and QCRI.


From dcard at andrew.cmu.edu  Wed Jul 30 16:58:19 2014
From: dcard at andrew.cmu.edu (Dallas Card)
Date: Wed, 30 Jul 2014 16:58:19 -0400
Subject: [CL+NLP Lunch] CL+NLP Lunch, Archna Bhatia, August 5 @ 2:00pm
Message-ID: <3d323ef9558cb2c9847697a79049cfe2.squirrel@webmail.andrew.cmu.edu>

Please join us for the next CL+NLP lunch at 2pm on August 5th, where
Archna Bhatia will be speaking about the cross-linguistic study of
form-function mappings. Lunch will be provided!

CL+NLP lunch <http://www.cs.cmu.edu/~nlp-lunch/>
Tuesday, August 5th at 2:00pm
GHC 6501

Archna Bhatia, LTI

*Studying form-function mappings cross-linguistically*

Abstract:
Languages carry form-function mappings, and while this property is shared
across languages, the exact same mappings are not. Cross-linguistic
studies of these mappings enable us to study the extent of systematicity
and the range of variation observed in the mappings across languages. This
can, in turn, shed some light on the nature of human language. In this
talk, I present our recent efforts to explore a phenomenon, definiteness,
which expresses the mapping between morphosyntax and the semantic,
pragmatic and discourse properties of noun phrases. Starting with a novel
language-independent annotation scheme for definiteness, and statistical
modeling of communicative functions, this work provides insight into the
form-function mappings of English noun phrases. Besides aiding in an
understanding of the nature of language in general, discoveries about the
form-function mapping across languages hold promise for various NLP
applications such as machine translation, knowledge base construction, and
information retrieval.

Bio:
Archna Bhatia is a postdoctoral researcher in the Language Technologies
Institute. She received her PhD in Linguistics from University of Illinois
at Urbana-Champaign in 2011. In her thesis research, she focussed on the
phenomenon of agreement in the context of coordination in Hindi and
compared it with two genetically unrelated languages, one with the same
and one with different word order properties. She is interested in
developing an understanding of the nature of language by applying
theoretical, descriptive, experimental and computational methods to study
its structure and acquisition. She has collaborated with researchers in
Linguistics, Computational Linguistics and Computer Science departments in
this endeavor.


-- 
Dallas Card
Machine Learning Department
Carnegie Mellon University





From dcard at andrew.cmu.edu  Tue Aug  5 09:51:59 2014
From: dcard at andrew.cmu.edu (Dallas Card)
Date: Tue, 5 Aug 2014 09:51:59 -0400
Subject: [CL+NLP Lunch] Reminder: CL+NLP Lunch, Archna Bhatia, TODAY @ 2:00pm
Message-ID: <24d697b8fde59e854996bae3f344aee3.squirrel@webmail.andrew.cmu.edu>

Please join us for the next CL+NLP lunch at 2pm on August 5th, where
Archna Bhatia will be speaking about the cross-linguistic study of
form-function mappings. Lunch will be provided!

CL+NLP lunch <http://www.cs.cmu.edu/~nlp-lunch/>
Tuesday, August 5th at 2:00pm
GHC 6501

Archna Bhatia, LTI

*Studying form-function mappings cross-linguistically*

Abstract:
Languages carry form-function mappings, and while this property is shared
across languages, the exact same mappings are not. Cross-linguistic
studies of these mappings enable us to study the extent of systematicity
and the range of variation observed in the mappings across languages. This
can, in turn, shed some light on the nature of human language. In this
talk, I present our recent efforts to explore a phenomenon, definiteness,
which expresses the mapping between morphosyntax and the semantic,
pragmatic and discourse properties of noun phrases. Starting with a novel
language-independent annotation scheme for definiteness, and statistical
modeling of communicative functions, this work provides insight into the
form-function mappings of English noun phrases. Besides aiding in an
understanding of the nature of language in general, discoveries about the
form-function mapping across languages hold promise for various NLP
applications such as machine translation, knowledge base construction, and
information retrieval.

Bio:
Archna Bhatia is a postdoctoral researcher in the Language Technologies
Institute. She received her PhD in Linguistics from University of Illinois
at Urbana-Champaign in 2011. In her thesis research, she focussed on the
phenomenon of agreement in the context of coordination in Hindi and
compared it with two genetically unrelated languages, one with the same
and one with different word order properties. She is interested in
developing an understanding of the nature of language by applying
theoretical, descriptive, experimental and computational methods to study
its structure and acquisition. She has collaborated with researchers in
Linguistics, Computational Linguistics and Computer Science departments in
this endeavor.


-- 
Dallas Card
Machine Learning Department
Carnegie Mellon University






From dcard at andrew.cmu.edu  Fri Aug  8 12:14:49 2014
From: dcard at andrew.cmu.edu (Dallas Card)
Date: Fri, 8 Aug 2014 12:14:49 -0400
Subject: [CL+NLP Lunch] CL+NLP Lunch, Kevin Gimpel, Friday August 15 @ 2:00pm
Message-ID: <8f554f7b9db39fe3101e7a6ed8d28050.squirrel@webmail.andrew.cmu.edu>

Please join us for the next CL+NLP lunch at 2pm on August 15th, where
Kevin Gimpel will be speaking about weakly-supervised NLP with
cost-augmented contrastive estimation. Lunch will be provided!

If you would like to set up a meeting with Kevin on August 15th, please
email Dallas Card <dcard at cs.cmu.edu>.

CL+NLP lunch <http://www.cs.cmu.edu/~nlp-lunch/>
Friday, August 15th at 2:00pm
GHC 8102

Kevin Gimpel, Toyota Technological Institute at Chicago

*Weakly-Supervised NLP with Cost-Augmented Contrastive Estimation*

Abstract:
Unsupervised NLP aims to discover meaningful structure in unannotated
text, such as parts-of-speech, morphological segmentation, or syntactic
structure.  Unsupervised systems improve when researchers incorporate
knowledge to bias learning to better capture characteristics of the
desired structure.  Contrastive estimation (CE; Smith and Eisner, 2005) is
a general approach to unsupervised learning with a particular way of
incorporating knowledge.  CE increases the likelihood of the observations
at the expense of those in a particular neighborhood of each observation. 
The neighborhood typically contains corrupted versions of the
observations.

In this talk, we generalize CE in two ways that allow us to add more
knowledge to unsupervised learning (thereby adding "weak" supervision). 
In particular, we augment CE with two types of cost functions, one on
observations and one on output structures.  The first allows the modeler
to specify not only the set of corrupted inputs for each observation, but
also how bad each one is.  The second lets us specify preferences on
desired output structures, regardless of the input sentence.  We evaluate
our approach, which we call cost-augmented contrastive estimation (CCE),
on unsupervised part-of-speech tagging of five languages from the PASCAL
2012 shared task.  We find that CCE improves over both standard CE and
strong benchmarks from the shared task.

This is joint work with Mohit Bansal.

Bio:
Kevin Gimpel is a research assistant professor at the Toyota Technological
Institute at Chicago, a philanthropically-endowed academic computer
science institute located on the campus of the University of Chicago.  In
2012, he received his PhD from the LTI where he was advised by Noah Smith.
 His recent research focuses on machine translation, unsupervised
learning, and structure prediction.


-- 
Dallas Card
Machine Learning Department
Carnegie Mellon University


From dcard at andrew.cmu.edu  Fri Sep 26 15:46:49 2014
From: dcard at andrew.cmu.edu (Dallas Card)
Date: Fri, 26 Sep 2014 15:46:49 -0400
Subject: [CL+NLP Lunch] CL+NLP Lunch, Kuzman Ganchev,
	Thursday October 2nd @ 1:00pm
Message-ID: <4d012869acc0dc687aa372502862081e.squirrel@webmail.andrew.cmu.edu>

Please join us for the next CL+NLP lunch at 1pm on October 2nd, where
Kuzman Ganchev will be speaking about frame semantic parsing. Lunch will
be provided!

CL+NLP lunch <http://www.cs.cmu.edu/~nlp-lunch/>
Thursday, October 2nd at 1:00pm
GHC 2109

Speaker: Kuzman Ganchev, Google Research

*Title: Some recent work on frame semantics*

Frame semantic parsing (sometimes called semantic role labeling) involves
identifying the predicate argument structure of a sentence. For  example
in the sentence "I want to run a marathon.", "want" and "run"  evoke
frames, and the frame for "run" has the agent "I" and patient "a 
marathon". The frame for "run" in that sentence will be different from the
frame for run in "I run the company". Frame semantic parsing is usually 
divided into frame identification and argument identification. I will 
present recent approaches from our group on each of these tasks: an 
embedding-based approach for frame identification, and a novel dynamic 
program for argument identification which allows computing posteriors  and
structured learning.

Kuzman Ganchev was born in Sofia, Bulgaria. He received his Ph.D. in 
computer and information science from the University of Pennsylvania, and 
has been working as a research scientist at Google since 2010, where his 
research focuses on machine learning applied to natural language
processing.


-- 
Dallas Card
Machine Learning Department
Carnegie Mellon University











From dcard at andrew.cmu.edu  Tue Sep 30 19:22:33 2014
From: dcard at andrew.cmu.edu (Dallas Card)
Date: Tue, 30 Sep 2014 19:22:33 -0400
Subject: [CL+NLP Lunch] Joint ML+NLP Lunch, Miguel Ballesteros,
 Monday October 6th @ 12:00pm
Message-ID: <035ef0f3e4dae84991cbfe07150c531a.squirrel@webmail.andrew.cmu.edu>

Please join us for a special joint ML+NLP lunch at noon on October 6th, where
Miguel Ballesteros will be speaking about dependency parsing. Lunch will
be provided!

ML+NLP lunch <http://www.cs.cmu.edu/~nlp-lunch/>
Monday, October 6th at 12:00pm
GHC 6115


Speaker: Miguel Ballesteros, Visiting Lecturer / Postdoc at Universitat
Pompeu Fabra

*Title: Going to the Roots of Dependency Parsing*

In this seminar I will first introduce transition-based dependency parsing
and present the
conclusions extracted from a journal paper that I have never had the
chance to present in public,
besides I'm going to sum up my current, past and future research
collaboration projects with some
new results and developments.
--
Dependency trees used in syntactic parsing often include a root node
representing a dummy
word prefixed or suffixed to the sentence, a device that is generally
considered a mere technical
convenience and is tacitly assumed to have no impact on empirical results.
We demonstrate that
this assumption is false and that the accuracy of data-driven dependency
parsers can in fact be
sensitive to the existence and placement of the dummy root node. In
particular, we show that
a greedy, left-to-right, arc-eager transition-based parser consistently
performs worse when the
dummy root node is placed at the beginning of the sentence (following the
current convention
in data-driven dependency parsing) than when it is placed at the end or
omitted completely.
Control experiments with an arc-standard transition-based parser and an
arc-factored graph-
based parser reveal no consistent preferences but nevertheless exhibit
considerable variation in
results depending on root placement. We conclude that the treatment of
dummy root nodes in
data-driven dependency parsing is an underestimated source of variation in
experiments and
may also be a parameter worth tuning for some parsers.


Miguel is a Visiting lecturer - Postdoc in Pompeu Fabra University,
Barcelona, Spain. He works on natural language processing and machine
learning with a special interest on linguistic structure prediction
problems, such as dependency parsing and phrase structure parsing.  He
completed his BsC,  MsC and PhD at the Universidad Complutense de Madrid.
During the last years, he was a Visiting Researcher in Universities of
Uppsala,  Birmingham and Singapore.


-- 
Dallas Card
Machine Learning Department
Carnegie Mellon University







From dcard at andrew.cmu.edu  Thu Oct  2 08:44:49 2014
From: dcard at andrew.cmu.edu (Dallas Card)
Date: Thu, 2 Oct 2014 08:44:49 -0400
Subject: [CL+NLP Lunch] Reminder: CL+NLP Lunch, Kuzman Ganchev,
	TODAY @ 1:00pm
Message-ID: <190f6785c694c76a6bd9d203f1332071.squirrel@webmail.andrew.cmu.edu>


CL+NLP lunch <http://www.cs.cmu.edu/~nlp-lunch/>
Thursday, October 2nd at 1:00pm
GHC 2109

Speaker: Kuzman Ganchev, Google Research

*Title: Some recent work on frame semantics*

Frame semantic parsing (sometimes called semantic role labeling) involves
identifying the predicate argument structure of a sentence. For  example
in the sentence "I want to run a marathon.", "want" and "run"  evoke
frames, and the frame for "run" has the agent "I" and patient "a
marathon". The frame for "run" in that sentence will be different from the
frame for run in "I run the company". Frame semantic parsing is usually
divided into frame identification and argument identification. I will
present recent approaches from our group on each of these tasks: an
embedding-based approach for frame identification, and a novel dynamic
program for argument identification which allows computing posteriors  and
structured learning.

Kuzman Ganchev was born in Sofia, Bulgaria. He received his Ph.D. in
computer and information science from the University of Pennsylvania, and
has been working as a research scientist at Google since 2010, where his
research focuses on machine learning applied to natural language
processing.


-- 
Dallas Card
Machine Learning Department
Carnegie Mellon University



From dcard at andrew.cmu.edu  Mon Oct  6 10:29:27 2014
From: dcard at andrew.cmu.edu (Dallas Card)
Date: Mon, 6 Oct 2014 10:29:27 -0400
Subject: [CL+NLP Lunch] Reminder: Joint ML+NLP Lunch, Miguel Ballesteros,
	TODAY @ 12:00pm
Message-ID: <6dcfa4d31ced9385389b694e492e692f.squirrel@webmail.andrew.cmu.edu>

Please join us for a special joint ML+NLP lunch at noon today, where
Miguel Ballesteros will be speaking about dependency parsing. Lunch will
be provided!

ML+NLP lunch <http://www.cs.cmu.edu/~nlp-lunch/>
Monday, October 6th at 12:00pm
GHC 6115


Speaker: Miguel Ballesteros, Visiting Lecturer / Postdoc at Universitat
Pompeu Fabra

*Title: Going to the Roots of Dependency Parsing*

In this seminar I will first introduce transition-based dependency parsing
and present the conclusions extracted from a journal paper that I have
never had the chance to present in public, besides I'm going to sum up my
current, past and future research collaboration projects with some new
results and developments.

--

Dependency trees used in syntactic parsing often include a root node
representing a dummy word prefixed or suffixed to the sentence, a device
that is generally considered a mere technical convenience and is tacitly
assumed to have no impact on empirical results. We demonstrate that this
assumption is false and that the accuracy of data-driven dependency
parsers can in fact be sensitive to the existence and placement of the
dummy root node. In particular, we show that a greedy, left-to-right,
arc-eager transition-based parser consistently performs worse when the
dummy root node is placed at the beginning of the sentence (following the
current convention in data-driven dependency parsing) than when it is
placed at the end or omitted completely. Control experiments with an
arc-standard transition-based parser and an arc-factored graph-based
parser reveal no consistent preferences but nevertheless exhibit
considerable variation in results depending on root placement. We conclude
that the treatment of dummy root nodes in data-driven dependency parsing
is an underestimated source of variation in experiments and may also be a
parameter worth tuning for some parsers.


Miguel is a Visiting lecturer - Postdoc in Pompeu Fabra University,
Barcelona, Spain. He works on natural language processing and machine
learning with a special interest on linguistic structure prediction
problems, such as dependency parsing and phrase structure parsing.  He
completed his BsC,  MsC and PhD at the Universidad Complutense de Madrid.
During the last years, he was a Visiting Researcher in Universities of
Uppsala,  Birmingham and Singapore.





From cdyer at cs.cmu.edu  Fri Oct 17 20:17:36 2014
From: cdyer at cs.cmu.edu (Chris Dyer)
Date: Fri, 17 Oct 2014 20:17:36 -0400
Subject: [CL+NLP Lunch] Sujith Ravi talk, Friday Oct 24, 11am (GHC 2109)
Message-ID: <CAEHEvxMMShE53ED4PHFuofrTOdWdr0BpHt_P4E3_DWTPVWrZuw@mail.gmail.com>

Sujith Ravi (Google) will be on campus next Friday Oct 24 and giving a
talk in GHC 2109 at 11am. Details below.

If you would like to meet with the speaker, please indicate your
availability here:
https://docs.google.com/document/d/1JjaXfGzt4Y__B1_ORJejOWgJaeGPSo56zdnd0dPFoxU/edit?usp=sharing

Title:  Large-scale Structure Prediction for Natural Language Processing

Abstract:

Natural language processing (NLP) systems have become ubiquitous for
data analysis in digital environments such as the Web and social
media. While great progress has been made in a wide range of areas,
building NLP systems from scratch still remains a daunting challenge
for many applications, especially when there is a need to target
different domains, languages or users. Current NLP systems heavily
rely on expensive human-annotated data and struggle to effectively
scale to the volume and characteristics of changing data environments,
complex modeling choices and wide range of applications. Overcoming
these challenges requires new advances in inference algorithms and
efficient approximate learning methods that reduce the computational
complexity involved in structured prediction problems.

In this talk, I will present a series of new powerful general-purpose
learning algorithms for large-scale structured prediction applicable
to a wide range of tasks in NLP, IR, speech and computer vision. This
work introduces novel algorithms for fast unsupervised and
semi-supervised learning that address current challenges and unlike
existing methods, the new approach scales to large data sizes and
dimensionality as well as complex structured models. The new
approaches fall under two major paradigms commonly used in machine
learning: ?probabilistic inference? and ?graph optimization?. This
talk will focus on the former---I will describe a new approach for
fitting mixtures of exponential families, which generalizes several
probabilistic models used in NLP and other areas. A major contribution
of our work is a novel sampling method that uses randomized techniques
like locality sensitive hashing to achieve high throughput in
generating proposals during sampling. This method scales very easily
to large data and model sizes achieving huge speedups of several
orders of magnitude over existing toolkits and outperform
state-of-the-art systems on a wide variety of structured prediction
tasks ranging from clustering to topic modeling to machine
translation. Moreover, we can efficiently parallelize the algorithm on
modern computing platforms to achieve even higher throughputs. In
addition, we also prove probabilistic error guarantees for the new
algorithm. These novel techniques show great promise for tackling
other complex AI problems such as deep language understanding and
building joint models of language and vision.

Bio: Sujith Ravi is a Research Scientist at Google since 2012. Prior
to that he was a Research Scientist at Yahoo! Research. He completed
his PhD at University of Southern California/Information Sciences
Institute. His main research interests span various problems and
theory related to the fields of Natural Language Processing (NLP) and
Machine Learning. He won the SIGKDD 2014 Best Research Paper Award and
a Best Paper Award nomination at ACL 2009. He is specifically
interested in large-scale unsupervised and semi-supervised methods and
their applications to structured prediction problems in NLP,
information extraction, multi-modal learning for language/vision, user
modeling in social media, graph optimization algorithms for
summarizing noisy data, computational decipherment and computational
advertising. He has published over 30 peer-reviewed papers in top-tier
conferences and journals. He was the organizer of the ICML-NAACL
symposium in 2013, Conference Workshop Co-Chair for NAACL-HLT 2013 and
serves on the PC for ACL, ICML, NIPS, NAACL, EMNLP, AAAI, KDD and
WSDM. His work has been reported in several magazines such as New
Scientist and ACM TechNews.

Homepage: http://www.sravi.org


From dcard at andrew.cmu.edu  Tue Oct 28 16:20:48 2014
From: dcard at andrew.cmu.edu (Dallas Card)
Date: Tue, 28 Oct 2014 16:20:48 -0400
Subject: [CL+NLP Lunch] CL+NLP Lunch, Wei Xu, Monday November 3rd @ 12:00pm
Message-ID: <77275ce31adb985072c81c7263bf489c.squirrel@webmail.andrew.cmu.edu>

Please join us for the next CL+NLP lunch at noon on Monday November 3rd,
where Wei Xu will be speaking about modeling paraphrase. Lunch will be
provided!

To arrange meetings with Wei on Monday, please contact Dallas Card
(dcard at cmu.edu).

---
ML+NLP lunch <http://www.cs.cmu.edu/~nlp-lunch/>
Monday, November 3rd at 12:00pm
GHC 6501

Speaker: Wei Xu, University of Pennsylvania

TITLE: Modeling Lexically Divergent Paraphrases in Twitter (and Shakespeare!)

ABSTRACT:
Paraphrases are alternative linguistic expressions of the same meaning.
Identifying paraphrases is fundamental to many natural language processing
tasks and has been extensively studied for the standard contemporary
English. In this talk I will present MULTIP (Multi-instance Learning
Paraphrase Model), a joint word-sentence alignment model suited to
identify paraphrases within the noisy user generated texts on Twitter. The
model infers latent word-level paraphrase anchors from only sentence level
annotations during learning. This is a major departure from previous
approaches that rely on lexical or distributional similarities over
sentence pairs. By reducing the dependence on word overlap as evidence of
paraphrase, our approach identifies more lexically divergent expressions
with equivalent meaning. For experiments, we constructed a Twitter
Paraphrase Corpus of about 19,000 sentences using a novel and efficient
crowdsourcing methodology. Our new approach improves the state-of-the-art
performance of a method that combines a latent space model with a
feature-based supervised classifier. I will also present findings on
paraphrasing between standard English and Shakespearean styles.

Joint work with Chris Callison-Burch (UPenn), Bill Dolan (MSR), Alan
Ritter (OSU), Yangfeng Ji (GaTech), Colin Cherry (NRC) and Ralph Grishman
(NYU).


Wei Xu is a postdoc in Computer and Information Science Department at
University of Pennsylvania, working with Chris Callison-Burch. Her
research focuses on paraphrases, social media and information extraction.
She received her PhD in Computer Science from New York University. She is
organizing the SemEval-2015 shared task on "Paraphrase and Semantic
Similarity in Twitter".  During her PhD, she visited University of
Washington for two years and interned at Microsoft Research, ETS and
Amazon.com.



From dcard at andrew.cmu.edu  Sun Nov  2 13:33:39 2014
From: dcard at andrew.cmu.edu (Dallas Card)
Date: Sun, 2 Nov 2014 13:33:39 -0500
Subject: [CL+NLP Lunch] Reminder: CL+NLP Lunch, Wei Xu,
	Monday November 3rd @ 12:00pm
Message-ID: <769d51634128e04fc13627b30b89991b.squirrel@webmail.andrew.cmu.edu>


Please join us for the next CL+NLP lunch at noon on Monday November 3rd,
where Wei Xu will be speaking about modeling paraphrase. Lunch will be
provided!

There are still time slots available to meet with Wei; if interested,
please contact Dallas Card (dcard at cmu.edu).

---
ML+NLP lunch <http://www.cs.cmu.edu/~nlp-lunch/>
Monday, November 3rd at 12:00pm
GHC 6501

Speaker: Wei Xu, University of Pennsylvania

TITLE: Modeling Lexically Divergent Paraphrases in Twitter (and Shakespeare!)

ABSTRACT:
Paraphrases are alternative linguistic expressions of the same meaning.
Identifying paraphrases is fundamental to many natural language processing
tasks and has been extensively studied for the standard contemporary
English. In this talk I will present MULTIP (Multi-instance Learning
Paraphrase Model), a joint word-sentence alignment model suited to
identify paraphrases within the noisy user generated texts on Twitter. The
model infers latent word-level paraphrase anchors from only sentence level
annotations during learning. This is a major departure from previous
approaches that rely on lexical or distributional similarities over
sentence pairs. By reducing the dependence on word overlap as evidence of
paraphrase, our approach identifies more lexically divergent expressions
with equivalent meaning. For experiments, we constructed a Twitter
Paraphrase Corpus of about 19,000 sentences using a novel and efficient
crowdsourcing methodology. Our new approach improves the state-of-the-art
performance of a method that combines a latent space model with a
feature-based supervised classifier. I will also present findings on
paraphrasing between standard English and Shakespearean styles.

Joint work with Chris Callison-Burch (UPenn), Bill Dolan (MSR), Alan
Ritter (OSU), Yangfeng Ji (GaTech), Colin Cherry (NRC) and Ralph Grishman
(NYU).


Wei Xu is a postdoc in Computer and Information Science Department at
University of Pennsylvania, working with Chris Callison-Burch. Her
research focuses on paraphrases, social media and information extraction.
She received her PhD in Computer Science from New York University. She is
organizing the SemEval-2015 shared task on "Paraphrase and Semantic
Similarity in Twitter".  During her PhD, she visited University of
Washington for two years and interned at Microsoft Research, ETS and
Amazon.com.