One core challenge in natural language research is to do robust, wide coverage semantic analysis. Recently, there has been progress towards solving this problem by developing algorithms for learning semantic parsers that map sentences to rich, logical representations of their meaning. State-of-the-art approaches have reached the level where they can, with sufficient training data, be used to learn highly accurate parsers for many different natural languages on a number of benchmark problems. However, the general applicability of this work has been limited by the focus on somewhat idealized conditions, where the application domain is of limited size, sentences are analyzed in isolation, and there is an exclusive focus on database query applications.
This CAREER project aims to build a framework for grounded semantic parsing that solves these challenges by reasoning about a sentence's possible meanings given its linguistic and situated context. This type of reasoning is necessary for extending existing learning approaches to fundamentally new applications, such as conversational understanding. However, it is also be crucial for the next generation of semantic parsers that learn from easily gathered data and scale to domains that are several orders of magnitude more complex than previously considered.
The project will extend the PI's educational and outreach efforts, including the creation of freely shared online content for introductory and advanced semantics topics. It will also enable new initiatives by the PI to increase diversity in computer science study and research, by supporting efforts to motivate students through early exposure to exciting language understanding problems.
