Xavier Carreras -- LSI/UPC

"Ramón y Cajal" researcher
NLP Research Group ∈ { TALP Research Centre, LSI Department } ∈ UPC

mylastname/ at/ lsi/upc/edu
phone : (+34) 93-413-7548
address : UPC Dept. LSI, Omega-317, c/Jordi Girona 1-3, 08034 Barcelona

xavier

I am a researcher at the Universitat Politècnica de Catalunya (UPC) since 2009, where I am a member of the NLP Research Group, the TALP Research Centre, and the LSI Department. From 2006 to 2009 I was a postdoc at MIT/CSAIL with Michael Collins. In 2005 I completed my PhD at the UPC; my advisor is Lluís Màrquez. My research is in Natural Language Processing and Machine Learning. I am specially interested in grammatical induction and parsing methods for syntactic-semantic analysis and translation of natural languages.

Activities

I participate in: XLike, the Pascal2 Network, FAUST, MOLTO, OpenMT-2, KNOW-2
I collaborate with dMetrics research.
I am secretary of SIGNLL (2011-2013), the ACL Special Interest Group on Natural Language Learning.
I am PC member of ACL-2013, NAACL-2013, CoNLL-2013, ...
I teach Advanced Natural Language Processing in the AI Masters program of FIB.

Past Activities

I was PC member of ACL-2012, EMNLP-CoNLL-2012, ECML/PKDD-2012, ICML-2011, IJCAI-2011, CoNLL-2011, EMNLP-2011, ACL-2010, ICML-2010, CoNLL-2010, EMNLP-2010 COLING-2010, ECAI-2010, NIPS-2009, EACL-2009, NAACL/HLT-2009, IJCAI-2009, ... and many past editions of the same conferences
I gave invited lectures at the 2011 and 2012 Lisbon Machine Learning School (LXMLS).
I co-organized the NIPS'12 Workshop on Cross-lingual Technologies (xLiTe)
I served as area chair for syntax and parsing for EACL 2012, with Kenji Sagae.
I served as area chair for syntax and parsing for ACL HLT 2011, with Ryan McDonald.
I co-chaired the demo track of ECML/PKDD 2010 with Ulf Brefeld.
I co-chaired CoNLL-2009 with Suzanne Stevenson
I was guest editor of Computational Linguistics : Special issue on Semantic Role Labeling (2008), with Lluís Màrquez, Ken Litkowski, and Suzanne Stevenson
In 2004 and 2005 I co-organized with Lluís Màrquez the CoNLL Shared Tasks on Semantic Role Labeling

Publications

PhD Thesis

Learning and Inference in Phrase Recognition: A Filtering-Ranking Architecture using Perceptron, Xavier Carreras, Ph.D Dissertation, Technical University of Catalonia, October 2005. [pdf] [ps.gz] [slides pdf] [bib] [code] [README file of the code]

Journal Articles

Exponentiated Gradient Algorithms for Conditional Random Fields and Max-Margin Markov Networks, Michael Collins, Amir Globerson, Terry Koo, Xavier Carreras and Peter Bartlett, JMLR, Volume 9, pages 1775-1822, 2008. [article]
Semantic Role Labeling: An Introduction to the Special Issue, Lluís Màrquez, Xavier Carreras, Ken Litkowski and Suzanne Stevenson, Computational Linguistics 34(2), pages 145-159, 2008. [article]
Combination Strategies for Semantic Role Labeling, Mihai Surdeanu, Lluís Màrquez, Xavier Carreras and Pere Comas, JAIR, Volume 29, pages 105-151, 2007. [article] [bib]
Filtering-Ranking Perceptron Learning for Partial Parsing, Xavier Carreras, Lluís Màrquez and Jorge Castro, Machine Learning Journal, Special Issue on Learning in Speech and Language Technologies, Volume 60, Issue 1-3, pages 41-71, Sept. 2005. [ps.gz] [pdf] [bib] [code] [README file of the code]
Margin maximization with feed-forward neural networks: a comparative study with SVM and AdaBoost, Enrique Romero, Lluís Màrquez and Xavier Carreras, Neurocomputing, 57(C):313-344, 2004. [pdf] [ps] [bib]
A Proposal for Wide-Coverage Spanish Named Entity Recognition, Montse Arévalo, Xavier Carreras, Lluís Màrquez, Toni Martí, Lluís Padró and Maria José Simon, SEPLN Journal "Procesamiento del Lenguaje Natural", Vol. 28. May 2002.
WordMed and Scriptum: Development of Terminological Resources for the Medical Practitioner, Victoria Arranz, Jordi Turmo, Xavier Carreras and Montserrat Arévalo, Terminology 7(1), John Benjamins Publishing Co., Amsterdam. 2001.

Papers in Conferences

Borja Balle, Ariadna Quattoni and Xavier Carreras. Local Loss Optimization in Operator Models: A New Insight into Spectral Learning. In Proceedings of ICML 2012. [pdf] [bib]
Franco M. Luque, Ariadna Quattoni, Borja Balle, and Xavier Carreras. Spectral Learning for Non-Deterministic Dependency Parsing. In Proceedings of EACL 2012, best paper award. [pdf] [bib]
Ariadna Quattoni, Xavier Carreras, and Antonio Torralba. A Latent Variable Ranking Model for Content-based Retrieval. In Proceedings of ECIR 2012. [pdf] [bib]
Borja Balle, Ariadna Quattoni, and Xavier Carreras. A Spectral Learning Algorithm for Finite State Transducers. In Proceedings of ECML-2011. [pdf] [bib] [poster pdf]
Xavier Carreras and Michael Collins. Non-projective Parsing for Statistical Machine Translation. In Proceedings of EMNLP-2009. [pdf] [bib] [slides]
Jun Suzuki, Hideki Isozaki, Xavier Carreras, and Michael Collins. An Empirical Study of Semi-supervised Structured Conditional Models for Dependency Parsing. In Proceedings of EMNLP-2009. [pdf] [bib]
Ariadna Quattoni, Xavier Carreras, Michael Collins, and Trevor Darrell. An Efficient Projection for L1,Infinity Regularization. In Proceedings of ICML-2009. [pdf] [bib]
Xavier Carreras, Michael Collins, and Terry Koo. TAG, Dynamic Programming, and the Perceptron for Efficient, Feature-rich Parsing. In Proceedings of CoNLL-2008, best paper award. [pdf] [bib] [slides]
Terry Koo, Xavier Carreras, and Michael Collins. Simple Semi-supervised Dependency Parsing. In Proceedings of ACL-2008. [pdf] [bib]
Amir Globerson, Terry Koo, Xavier Carreras, and Michael Collins. Exponentiated Gradient Algorithms for Log-Linear Structured Prediction. In Proceedings of ICML 2007. [pdf] [bib]
Terry Koo, Amir Globerson, Xavier Carreras, and Michael Collins. Structured Prediction Models via the Matrix-Tree Theorem. In Proceedings of EMNLP-CoNLL 2007. [pdf] [bib]
Xavier Carreras. Experiments with a Higher-Order Projective Dependency Parser. In Proceedings of the EMNLP-CoNLL 2007 Shared Task. [pdf] [ps] [bib]
Projective Dependency Parsing with Perceptron, Xavier Carreras, Mihai Surdeanu, and Lluís Màrquez, Proceedings of the CoNLL-X Shared Task, New York 2006. [pdf] [ps] [slides in pdf]
Introduction to the CoNLL-2005 Shared Task: Semantic Role Labeling, Xavier Carreras and Lluís Màrquez, In Proceedings of the CoNLL-2005 Shared Task, Ann Arbor, MI USA. June 2005. [pdf] [bib], See also the webpage of CoNLL Shared Tasks on SRL, 2004 and 2005 editions.
Exploiting Diversity of Margin-based Classifiers, Enrique Romero, Xavier Carreras, and Lluís Màrquez, In Proceedings of the 2004 International Joint Conference on Neural Networks (IJCNN), Budapest, Hungary, July 2004. [pdf]
Introduction to the CoNLL-2004 Shared Task: Semantic Role Labeling, Xavier Carreras and Lluís Màrquez, In Proceedings of the CoNLL-2004 Shared Task, Boston, MA USA. May 2004. [ps] [ps.gz] [pdf]
Hierarchical Recognition of Propositional Arguments with Perceptrons, Xavier Carreras, Lluís Màrquez, and Grzegorz Chrupała, In Proceedings of the CoNLL-2004 Shared Task, Boston, MA USA. May 2004. [pdf]
FreeLing: An Open-Source Suite of Language Analyzers, Xavier Carreras, Isaac Chao, Lluís Padró, and Muntsa Padró, In Proceedings of the 4th International Conference on Language Resources and Evaluation (LREC'04). Lisbon, Portugal. 2004. [pdf] [FreeLing homepage]
Online Learning via Global Feedback for Phrase Recognition, Xavier Carreras and Lluís Màrquez, NIPS-2003, Vancouver, Canada. December 2003. [ps]
Phrase Recognition by Filtering and Ranking with Perceptrons, Xavier Carreras and Lluís Màrquez, RANLP-2003, Borovets, Bulgaria. September 2003. [ps.gz] [pdf] [slides pdf]
A Simple Named Entity Extractor Using AdaBoost, Xavier Carreras, Lluís Màrquez, and Lluís Padró, Proceedings of the CoNLL-2003 Shared Task. Edmonton, Canada. June 2003. [ps]
Learning a Perceptron-Based Named Entity Chunker via Online Recognition Feedback, Xavier Carreras, Lluís Màrquez, and Lluís Padró, Proceedings of the CoNLL-2003 Shared Task. Edmonton, Canada. June 2003. [ps]
Low-cost Named Entity Classification for Catalan: Exploiting Multilingual Resources and Unlabeled Data, Lluís Màrquez, Adrià de Gispert, Xavier Carreras, and Lluís Padró, 1st ACL Workshop on Multilingual and Mixed-language Named Entity Recognition: Combining Statistical and Symbolic Models. Sapporo, Japan. July 2003. [ps] [pdf]
Named Entity Recognition For Catalan Using Spanish Resources, Xavier Carreras, Lluís Màrquez, and Lluís Padró, 10th Conference of the European Chapter of the Association for Computational Linguistics (EACL'03). Budapest, Hungary. April 2003. [ps] [pdf]
Named Entity Extraction Using AdaBoost, Xavier Carreras, Lluís Màrquez, and Lluís Padró, In Proceedings of the CoNLL-2002 Shared Task, Taipei, Taiwan, September 2002. [ps.gz] [pdf] [slides ps]
Learning and Inference for Clause Identification, Xavier Carreras, Lluís Màrquez, Vasin Punyakanok, and Dan Roth, 13th European Conference on Machine Learning (ECML'02). Helsinki, Finland. August 2002. [ps.gz] [pdf] [pdf slides]
Wide-Coverage Spanish Named Entity Extraction, Xavier Carreras, Lluís Màrquez, and Lluís Padró, VIII Conferencia Iberoamericana de Inteligencia Artificial, IBERAMIA'02. Sevilla, Spain. November 2002. [ps.gz]
A Flexible Distributed Architecture for Natural Language Analyzers, Xavier Carreras and Lluís Padró, Conference on Language Resources and Evaluation (LREC'02). Las Palmas de Gran Canaria, Spain. 2002. [ps.gz]
Boosting Trees for Clause Splitting, Xavier Carreras and Lluís Màrquez, In Proceedings of the CoNLL-2001 Shared Task. Toulouse, France. 2001. [ps] [bib] [talk slides]
Boosting Trees for Anti-Spam Email Filtering, Xavier Carreras and Lluís Màrquez, Conference on Recent Advances in NLP (RANLP'01). Tzigov Chark, Bulgaria. 2001. [ps] [bib] [ps slides]

Software

egstra: Exponentiated-Gradient algorithms for STRuctured prediction. This C++ code implements EG algorithms to train max-margin and log-linear models for structured prediction tasks. The package includes other popular learning algorithms as well, namely Perceptron and SGD. The current package includes algorithms for first-order dependency parsing.
boostree : A C++ implementation of Schapire and Singer's AdaBoost.MH, with decision-tree weak learning. The algorithm is suited for multi-label classification problems, and sparse example representations based on binary-valued features (typical for NLP problems). This code was used to develop many CoNLL shared task systems which obtained state-of-the-art-results. Check the documentation of the package.

Last modified: Feb. 2013