3/11/2021

Carlos Gómez Rodríguez and Daniel Fernández González will attend the 2021 Conference on Empirical Methods in Natural Language Processing that will take place in Punta Cana (Dominican Republic) from 7 to 11 November 2021. They will present their developments at the research group:

  • Reducing Discontinuous to Continuous Parsing with Pointer Network Reordering

Discontinuous parsing is an especially challenging form of syntactic parsing which is needed to describe various syntactic phenomena, especially on free-word-order languages, but is often avoided because it makes parsing difficult. This paper presents an approach to reduce discontinuous to continuous (standard) parsing, in such a way that we can use any existing continuous parser to recognize those discontinuous phenomena, thus making their handling much easier. The approach is based on reordering sentences with a specific neural network called a Pointer Network before and after parsing.