Adaptive Algorithms for Semantic Dependency Parsing
Research project funded by a Google Faculty Research Award, 2016
The purpose of this project is to design, analyze, and evaluate new, powerful algorithms for semantic dependency parsing, the problem of mapping a natural language sentence into a graph expressing the predicate–argument relationships for all content words. Semantic dependency parsing is currently attracting interest in natural language processing because of the wide range of its applications and the increased availability of linguistic data that can be used to induce parsers using machine learning. At the same time, the existing repertoire of practical algorithms for semantic dependency parsing is incomplete. In particular, we see a need for algorithms that can learn to trade off between the conflicting interests of high coverage on the data on the one hand and high expressiveness in the predictive models on the other hand. By developing such algorithms, this project will advance the state of the art in semantic dependency parsing, and lead to a deeper understanding of the associated computational problems.
Page responsible: Marco Kuhlmann
Last updated: 2018-05-18