Automatically Deriving Adjectival Scales from Neural Network Language Models - 2013

Watch a 57-minute lecture from the Center for Language & Speech Processing at Johns Hopkins University where Marie-Catherine de Marneffe explores the automatic derivation of orderings and meanings of gradable adjectives. Learn how to leverage web-based text data and neural network language models to determine scalar relationships between modifiers like "wonderful" and "worth seeing." Discover techniques for extracting continuous space word representations to create high-quality adjectival scales, with applications in both general language understanding and specialized domains like biomedical text analysis. Gain insights into computational linguistic methods that capture meaning beyond literal interpretations, drawing from de Marneffe's expertise in developing practical representations of grammatical relations and predicate argument structure.