Alexander Koller

The Constraint Language for Lambda Structures

This paper presents the Constraint Language for Lambda Structures(CLLS), a first-order language for semantic underspecification thatconservatively extends dominance constraints. It is interpreted overlambda structures, tree-like structures that encode λ-terms. Based onCLLS, we present an underspecified, uniform analysis of scope,ellipsis, anaphora, and their interactions. CLLS solves a variablecapturing problem that is omnipresent in scope underspecification andcan be processed efficiently.

Generation as Dependency Parsing

Natural-Language Generation from flat semantics is an NP-complete problem. This makes it necessary to develop algorithms that run with reasonable efficiency in practice despite the high worst-case complexity. We show how to convert TAG generation problems into dependency parsing problems, which is useful because optimizations in recent dependency parsers based on constraint programming tackle exactly the combinatorics that make generation hard. Indeed, initial experiments display promising runtimes.

Dominance Constraints: Algorithms and Complexity

Dominance constraints for finite tree structures are widely used in several areas of computational linguistics including syntax, semantics, and discourse. In this paper, we investigate algorithmic and complexity questions for dominance constraints and their first-order theory. The main result of this paper is that the satisfiability problem of dominance constraints is NP-complete. We present two NP algorithms for solving dominance constraints, which have been implemented in the concurrent constraint programming language Oz. Despite the intractability result, the more sophisticated of our algorithms performs well in an application to scope underspecification. We also show that the positive existential fragment of the first-order theory of dominance constraints is NP-complete and that the full first-order theory has non-elementary complexity.

Computing Locally Coherent Discourses

We present the first algorithm that computes optimal orderings of sentences into a locally coherent discourse. The algorithm runs very efficiently on a variety of coherence measures from the literature. We also show that the discourse ordering problem is NP-complete and cannot be approximated.

An efficient graph algorithm for dominance constraints

Dominance constraints are logical descriptions of trees that are widely used in computational linguistics. Their general satisfiability problem is known to be NP-complete. Here we identify normal dominance constraints and present an efficient graph algorithm for testing their satisfiability in deterministic polynomial time. Previously, no polynomial time algorithm was known.

Experiences with Planning for Natural Language Generation

Natural language generation (NLG) is a major subfield of computational linguistics with a long tradition as an application area of automated planning systems. While current mainstream approaches have largely ignored the planning approach to NLG, several recent publications have sparked a renewed interest in this area. In this article, we investigate the extent to which these new NLG approaches profit from the advances in planner expressiveness and efficiency. Our findings are mixed. While modern planners can readily handle the search problems that arise in our NLG experiments, their overall runtime is often dominated by the grounding step they perform as preprocessing. Furthermore, small changes in the structure of a domain can significantly shift the balance between search and preprocessing. Overall, our experiments show that the off‐the‐shelf planners we tested are unusably slow for nontrivial NLG problem instances. As a result, we offer our domains and experiences as challenges for the planning community.

The first challenge on generating instructions in virtual environments

This paper describes the First Challenge on Generating Instructions in Virtual Environments (GIVE-1). GIVE is a shared task for generation systems which give real-time natural-language instructions to users in a virtual 3D world. These systems are evaluated by connecting users and NLG systems over the Internet. We describe the design and results of GIVE-1 as well as the participating NLG systems, and validate the experimental methodology by comparing the results collected over the Internet with results from a more traditional laboratory-based experiment.

Referring expressions as formulas of description logic

In this paper, we propose to reinterpret the problem of generating referring expressions (GRE) as the problem of computing a formula in a description logic that is only satisfied by the referent. This view offers a new unifying perspective under which existing GRE algorithms can be compared. We also show that by applying existing algorithms for computing simulation classes in description logic, we can obtain extremely efficient algorithms for relational referring expressions without any danger of running into infinite regress.

A relational syntax-semantics interface based on dependency grammar

We propose a syntax-semantics interface that realises the mapping between syntax and semantics as a relation and does not make functionality assumptions in either direction. This interface is stated in terms of Extensible Dependency Grammar (XDG), a grammar formalism we newly specify. XDGs constraint-based parser supports the concurrent flow of information between any two levels of linguistic representation, even when only partial analyses are available. This generalises the concept of underspecification.

Sentence generation as a planning problem

We translate sentence generation from TAG grammars with semantic and pragmatic information into a planning problem by encoding the contribution of each word declaratively and explicitly. This allows us to exploit the performance of off-the-shelf planners. It also opens up new perspectives on referring expression generation and the relationship between language and action.