Efficient Algorithms in Computational Linguistics (Former Research Group)

The research group "Efficient Algorithms in Computational Linguistics", headed by Alexander Koller, was part of the Cluster of Excellence MMCI from August 2008 until March 2011.

Vision and Research Strategy

The current state of the art in computational linguistics is characterized by a contrast between “deep” approaches, which attempt to compute and work with linguistically satisfying representations, and “shallow” approaches, which focus on automatically learning useful regularities from large data sets. Deep approaches are currently limited by computational inefficiency and a lack of necessary knowledge resources. The aim of this research group is to address these issues and help make deep processing more realistic.

Our research focuses on three areas. First, we develop algorithms for dealing with semantic ambiguity and learning knowledge resources that are necessary for computing semantic inferences. Second, we investigate formal relationships between different grammar formalisms and work on parsing algorithms. Third, we exploit the expressive power of deep approaches to dialogue and natural language generation in order to develop models of communication in physical environments. A defining feature of our research is to identify and exploit connections between problems in natural language processing and other areas of computer science, including formal language theory, AI, and bioinformatics.

Composition of Group

The research group Efficient Algorithms in Computational Linguistics was founded in August 2008. The group had two members in addition to the group head, Alexander Koller: Konstantina Garoufi joined as a PhD student in December 2008, and Dr. Andrew Gargett joined as a postdoc in January 2009.

Research Topics and Achievements

Projects and Collaborations

Prizes and Awards

Dr. Alexander Koller

Dr. Alexander Koller

Since April 2011, Alexander Koller is a full professor of Theoretical Computational Linguistics, University of Potsdam.

Fon: +49 331 977 2692

Publications