Constructing Tunable Sentence Simplification Models using Deep Learning
Sanqiang Zhao · D-Scholarship@Pitt (University of Pittsburgh) · 2021
Sentence simplification aims to reduce the complexity of a sentence while retaining its original meaning so that certain individuals can read and understand it. Substitution, Dropping, Reordering, and Splitting are widely accepted as four important operations. Recent approaches view the simplification process as a monolingual text-to-text translation, where the translation model learns the operations automatically from examples of complex-simplified sentence pairs extracted from online resources. In the current literature, the two publicly available resources commonly used are Wikipedia and Newsela. However, both resources are limited in several ways, and only contribute to certain operations.