Scope and Negation:
Typological Diversity Meets Computational Semantics

Instructor(s): Gerald Penn and Frank Richter

A linguistic account of sentential negation minimally comprises a syntactic component, a logical representation, and a notion of semantic inference. While linguists often postulate elaborate abstract representations, this community has not drawn an explicit connection to genuine semantical inference. Researchers in NLP, on the other hand, opt for shallower representations of semantic content that negation simply defies.

This course aims to advance our understanding of negation by reappraising recent developments in both camps. After a survey of various manifestations of negation (negative concord, NPI licensing, interaction with modals), we present a logic-based architecture to describe them. We introduce techniques for automated theorem proving and contrast the domains of inference relative to real-world knowledge and of inference relative to logical-form composition. We cover contributions to the Pascal RTE challenges as examples of how semantic processing is being combined with linguistic analysis, and discuss how inference can inform and guide logical-form composition.

A basic familiarity with semantics, logic and syntactic theory. No particular familiarity with theorem proving, natural language processing or current theories of scope and negation required.

Course ID:

Mon & Thu 10:30-12:15

Classroom: ECCE 152

Areas of Linguistics:
Computational Linguistics
Semantics, Pragmatics and Discourse