Computational Linguistics Colloquium

Thursday, January 24, 16:15, Building 17, Seminar Room

The Structure and Interpretation of Discourse: Evidence from Ellipsis

Beginning with early work such as Sag (76), the interpretation of ellipsis has been an enduring preoccupation of linguistics research. According to Sag's account, the facts of ellipsis reflect logical aspects of the representation of pronouns, which are ambiguous between a bound and referential reading. On this influential view, ellipsis reveals an intricate machinery of lambda binding, pronoun indexing, and scope relations. Subsequent research has shed doubt on this pleasing picture -- it has been shown that ellipsis interpretation is somewhat less well-behaved than Sag's theory permits. At the same time, while Sag's theory was developed in a framework that treated the logical form of sentences in isolation, subsequent research has lead to a variety of perspectives which can encompass the interpretation of multi-sentence discourse.

In this talk I will pursue this shift in perspective, and suggest that ellipsis is revealing in a rather different sense than suggested by Sag: instead of providing insight into the intricacies of the logical form of sentences, ellipsis might be a window onto the general mechanisms governing the structure and interpretation of multi-sentence discourse. I will present a variety of evidence that ellipsis interpretation is strongly constrained by discourse structure; in particular, I will argue that parallelism constraints must be applied in accordance with discourse structure. I will consider two approaches to inter-sentential interpretation: the Etype approach and DRT. I will show that, while both approaches permit solutions to some empirical problems alluded to above, they each have their own problems, and I will conclude by sketching a new approach which combines virtues of both the Etype account and DRT.

If you would like to meet with the speaker, please contact Katrin Erk.