Controllable Sentence Simplification via Operation Classification

Abstract

Different types of transformations have been used to model sentence simplification ranging from mainly local operations such as phrasal or lexical rewriting, deletion and re-ordering to the more global affecting the whole input sentence such as sentence rephrasing, copying and splitting. In this paper, we propose a novel approach to sentence simplification which encompasses four global operations: whether to rephrase or copy and whether to split based on syntactic or discourse structure. We create a novel dataset that can be used to train highly accurate classification systems for these four operations. We propose a controllable-simplification model that tailors simplifications to these operations and show that it outperforms both end-to-end, noncontrollable approaches and previous controllable approaches.

Simplicity Level Estimate (SLE): A Learned Reference-Less Metric for Sentence Simplification

“…This could be useful in the evaluation of controllable simplification systems, which should be able to satisfy simplification requirements of specific user groups or reading levels (Martin et al, 2020;Cripwell et al, 2022;Yanamoto et al, 2022). As it is trained with aggregate document-level accuracy in mind, SLE could also possibly be used to evaluate document simplification -either via averaging sentence scores or using some other aggregation method.…”

A Review of Research-based Automatic Text Simplification Tools

“…In this section, we analyse the approach taken for text simplification. In general, the automatic simplification process comprises two stages (Cripwell et al, 2023): (1) the simplification plan, which refers to the decision about what linguistic aspect to simplify, for instance, identifying complex words or sentences; and (2) the simplification stage, when the plan to produce the simplified content is applied, e.g., splitting long sentences. It is worth noting that a system may perform these tasks holistically without a clear distinction between stages, as in neural generative models (Ondov et al, 2022).…”

Document-Level Planning for Text Simplification

“…These show some capacity for general improvement over end-to-end systems, while also significantly improving performance for specific operations (e.g. splitting in the case of Cripwell et al (2022)).…”

“…In response, controllable simplification systems have been proposed which either constrain attributes of the output (length, amount of paraphrasing, lexical and syntactic complexity) (Martin et al, 2020) or explicitly specify which simplification operation to perform (Alva-Manchego et al, 2017;Dong et al, 2019;Malmi et al, 2019;Scarton et al, 2020;Maddela et al, 2021;Cripwell et al, 2022).…”

“…Applying a binary classifier to predict whether simplification should be performed has been found to improve SARI results, reducing conservatism and spurious transformations (Scarton et al, 2020;Garbacea et al, 2021). Others have proposed multi-class systems to predict sentence-level operations that are then used to condition a generative model (Scarton and Specia, 2018;Scarton et al, 2020;Cripwell et al, 2022). These show some capacity for general improvement over end-to-end systems, while also significantly improving performance for specific operations (e.g.…”