IWPT 2011 - Invited Speakers

Invited Speakers

Ina Bornkessel-Schlesewsky (Philipps-Universität Marburg, Germany)
Michael Collins (Columbia University, USA)
Mark Steedman (University of Edinburgh, UK)

Talk Abstracts

Towards a Neurobiologically Plausible Model of Human Sentence Comprehension across Languages
Ina Bornkessel-Schlesewsky (Philipps-Universität Marburg, Germany)

Among human cognitive abilities, language is singular in the diversity of its manifestations: over 6000 languages are spoken in the world today. Some of the major challenges in modelling how language is processed by the human brain thus lie in explaining (a) how this diversity is handled, and (b) whether there are nevertheless some underlying generalisations that recur across languages of different types. Furthermore, an adequate model should be neurobiologically plausible, i.e., respect what we know about the structure and function of the human brain. In this presentation, I will describe a line of research in which we have attempted to take up these challenges at the level of sentence comprehension. Based on the results of neurophysiological experiments in a range of typologically varied languages, I will argue for a comprehension architecture that is actor-centred, i.e., focused on identifying the participant primarily responsible for the state of affairs under discussion. I will introduce the latest version of a comprehension model (extended Argument Dependency Model, eADM; Bornkessel & Schlesewsky, 2006), the architecture of which is built around actor-centrality as a design principle, and will describe how it accounts for potential universals of comprehension and critical dimensions of variation.

Lagrangian Relaxation for Inference in Natural Language Processing
Michael Collins (Columbia University, USA)

There has been a long history in combinatorial optimization of methods that exploit structure in complex problems, using methods such as dual decomposition or Lagrangian relaxation. These methods leverage the observation that complex inference problems can often be decomposed into efficiently solvable sub-problems. Thus far, however, these methods are not widely used in NLP.

In this talk I'll describe recent work on inference algorithms for NLP based on Lagrangian relaxation. In the first part of the talk I'll describe work on non-projective parsing. In the second part of the talk I'll describe an exact decoding algorithm for syntax-based statistical translation. If time permits, I'll also briefly describe algorithms for dynamic programming intersections (e.g., the intersection of a PCFG and an HMM), and for phrase-based translation.

For all of the problems that we consider, the resulting algorithms produce exact solutions, with certificates of optimality, on the vast majority of examples; the algorithms are efficient for problems that are either NP-hard (as is the case for non-projective parsing, or for phrase-based translation), or for problems that are solvable in polynomial time using dynamic programming, but where the traditional exact algorithms are far too expensive to be practical.

While the focus of this talk is on NLP problems, there are close connections to inference methods, in particular belief propagation, for graphical models. Our work was inspired by recent work that has used dual decomposition as an alternative to belief propagation in Markov random fields.

This is joint work with Yin-Wen Chang, Tommi Jaakkola, Terry Koo, Sasha Rush, and David Sontag.

Computing Scope in a CCG Parser
Mark Steedman (University of Edinburgh, UK)

Ambiguities arising from alternations of scope in interpretations for multiply quantified sentences appear to require grammatical operations that compromise the strong assumptions of syntactic/semantic transparency and monotonicity underlying the Frege-Montague approach to the theory of grammar. Examples that have been proposed include covert movement at the level of logical form, abstraction or storage mechanisms, and proliferating type-changing operations. The paper examines some interactions of scope alternation with syntactic phenomena including coordination, binding, and relativization. Starting from the assumption of Fodor and Sag, and others, that many expressions that have been treated as generalized quantifiers are in fact referential expressions, and using Combinatory Categorial Grammar (CCG) as a grammatical framework, the paper presents an account of quantifier scope ambiguities according to which the available readings are projected directly from the lexicon by the combinatorics of the syntactic derivation, without any independent manipulation of logical form and without recourse to otherwise unmotivated type-changing operations. As a direct result, scope ambiguity can be efficiently processed using packed representations from which the available readings can be simply enumerated.