Abstract

At this session of the NLIP Reading Group we’ll be discussing (part of) the following paper:

Alexander Koller, Stefan Thater, and Manfred Pinkal. Scope underspecification with tree descriptions: Theory and practice. To appear in Siekmann, J., Crocker, M. (eds.), Resource Adaptive Cognitive Processes, Berlin, Springer.

Abstract: In this paper, we outline the results of the CHORUS project on the topic of semantic underspecification, along two major lines of research: We review a series of solvers for dominance constraints and dominance graphs, two closely related underspecification formalisms developed in CHORUS , and present research results concerned with semantic construction and the development of wide-coverage underspecification methods that can be applied on “real world” corpus data.

Reading Notes Over the last ten-or-so years, Koller & Thater have published a series of papers on the theoretical and computational properties of semantic representations like MRS w.r.t. the problem of scope underspecification.

This paper makes for a really nice read for a general-interest NLP audience, because it summarizes this work concisely. The paper is 27 book-sized pages, if you want to skip stuff, Richard suggests skipping the following sections: Section 1 (intro), Section 2.3 (theoretic subtlety), Sections 3.1/3.2 (historical), Section 4.1 (if you already know MRS ), Section 6 (recap).