The next talk in our 2025-2026 McGill Linguistics Colloquium Series will be given by Simon Charlow (Yale University) on Friday, October 24th at 3:30pm at Leacock 232. The details of the talk are given below.
Title: Effect-oriented interpreters for natural language
Abstract: Computer programs are often factored into ‘pure’ components — simple, total functions from inputs to outputs — and components that incur ‘side effects’ such as errors, changes to memory, parallel threads, abortion of the current command, etc. Drawing on recent work with Dylan Bumford (UCLA), this talk makes the case that human languages are similarly organized around the give and pull of pure and effectful processes and shows how denotational techniques from computer science can be leveraged to support elegant and illuminating semantic analyses of natural language. Taking this perspective helps us come to grips with the otherwise bewilderingly rich multi-dimensionality of linguistic meaning, from definiteness and context-sensitivity to indefiniteness and disjunction, from anaphora and apposition to quantification, focus, scope, and evaluation order.
Previously, effect-oriented approaches to semantics have been operationalized via type-shifters, scope-shifting, and silent elements in the syntax. The talk will point out some unsatisfactory aspects of this state of affairs, and motivate a new approach, in which the linguistic interpreter is modelled as a kind of effect-driven compiler. I’ll map out a conservative extension of standard techniques for ‘type-driven translation’ with resources that allow a linguistic interpreter to generate and handle arbitrary permutations of effects, and their interactions. We’ll use a computational implementation of these ideas to interactively generate analyses and explore predictions.