Ivan Titov


ivan.titov   (at)  cui.unige.ch

CS PhD Student
Geneva Artificial Intelligence Lab

Address

Center Universitaire d'Informatique
University of Geneva
Battelle bâtiment A
7, route de Drize
CH-1227 Carouge (Geneva)
Switzerland

Phone: +41 22 379 02 15
Fax: +41 22 379 02 50


About

I am a PhD student in the Computer Science department at University of Geneva. My interests are in natural language processing, machine learning and computational complexity theory. My advisors are James Henderson at University of Edinburgh and Christian Pellegrini at University of Geneva.

February 2008: I moved to the Cognitive Computation Group at the University of Illinois at Urbana-Champaign. You can contact me at titov   (at)  uiuc.edu

[Download CV]


Refereed Publications

2008

2007

2006

2005
Other Papers

Research Projects

  • Neural Networks for Structure Processing Applied to Broad Coverage Natural Language Parsing (finished: Oct, 2005)
  • Kernel-Based Structure Processing for Natural Language Parsing (Nov, 2005 - ...)


Education


Master Thesis

Complexity and Approximation of Optimal Communication Cost Spanning Tree Problems, 2003
Under supervision of Nikolay N. Vasiliev, St. Petersburg Department of Steklov Institute of Mathematics and St. Petersburg State Polytechnical University).


Software

ISBN Dependency Parser

New! ISBN parser described in [Titov and Henderson, IWPT 2007] and evaluated in [Titov and Henderson, EMNLP-CoNLL 2007].

[Download it here]

ISBN Constituent Parser

ISBN parser described in [Titov and Henderson, ACL 2007, Titov and Henderson, ICML 2007] will be available for download very soon.

SSN Parsing Package

This package implements a statistical syntactic parser for natural language sentences. It estimates the probability of a parse using a neural network architecture for structure processing called Simple Synchrony Networks. This package includes two trained parsers which can be used to parse English text. It also supports training new generative models or new conditionally-trained models. For a description of the generative SSN parsing model, see [Henderson, NAACL 2003]. For a description of conditional training with this model see [Henderson, ACL 2004]. A future version of this package will include the code to support deriving data-defined kernels from SSNs [Henderson and Titov, ACL 2005], [Titov and Henderson, IJCNN 2005]. This software was originally written by James Henderson and modified substantially by Ivan Titov.

[Download it here]


Teaching

  • Pattern Recognition
  • Algorithms for Learning and Optimization