Cognitive Simplification Operations Improve Text Simplification
41638Authors
nan
Abstract
AbstractText Simplification (TS) is the task of converting a text into a form that is easier to read while maintaining the meaning of the original text. A sub-task of TS is Cognitive Simplification (CS), converting text to a form that is readily understood by people with cognitive disabilities without rendering it childish or simplistic. This sub-task has yet to be explored with neural methods in NLP, and resources for it are scarcely available. In this paper, we present a method for incorporating knowledge from the cognitive accessibility domain into a TS model, by introducing an inductive bias regarding what simplification operations to use. We show that by adding this inductive bias to a TS-trained model, it is able to adapt better to CS without ever seeing CS data, and outperform a baseline model on a traditional TS benchmark. In addition, we provide a novel test dataset for CS, and analyze the differences between CS corpora and existing TS corpora, in terms of how simplification operations are applied.
Keywords
Full Content
Proceedings of the 26th Conference on Computational Natural Language Learning (CoNLL), pages 241 - 265 December 7-8, 2022 ©2022 Association for Computational Linguistics Cognitive Simplification Operations Improve Text Simplification Eytan Chamovitz Department of Computer Science Tel Aviv University eytanc@gmail.com Omri Abend Department of Computer Science Hebrew University of Jerusalem omri.abend@mail.huji.ac.il Abstract Text Simplification (TS) is the task of convert- ing a text into a form that is easier to read while maintaining the meaning of the original text. A sub-task of TS is Cognitive Simplification (CS), converting text to a form that is readily understood by people with cognitive disabili- ties without rendering it childish or simplistic. This sub-task has yet to be explored with neu- ral methods in NLP, and resources for it are scarcely available. In this paper, we present a method for incorporating knowledge from the cognitive accessibility domain into a TS model, by introducing an inductive bias regarding what simplification operations to use. We show that by adding this inductive bias to a TS-trained model, it is able to adapt better to CS without ever seeing CS data, and outperform a baseline model on a traditional TS benchmark. In ad- dition, we provide a novel test dataset for CS, and analyze the differences between CS cor- pora and existing TS corpora, in terms of how simplification operations are applied. 1 Introduction Text Simplification (TS) is the task of converting text into a form that is easier to understand by modifying its syntax and/or the words used in it, while maintaining the original text’s meaning (Alva-Manchego et al., 2020b). TS is a very diverse task that can include sim- plifications aimed at different target audiences. TS is often operationalized in NLP using a number of particular corpora to train and evaluate neural models (see §3), whose target audiences are mostly second language learners, primary school students or adults with learning disabilities. For brevity, this paper will refer by TS to this concrete formulation of the simplification task, rather than the abstract, general notion of the task of simplifying text. Cognitive Simplification (CS) is the task of con- verting text to a form that is clear, simple, and read- ily understood by people with cognitive disabilities (Yalon-Chamovitz, 2009; Yalon-Chamovitz et al., 2016; Yalon-Chamovitz and Avidan-Ziv, 2016).1 The procedure includes structural and lexical mod- ifications that reduce the text’s complexity, while preserving as much of the meaning and informa- tion content as possible, and without rendering it childish or simplistic. See Figure 1. The following example illustrates the differences and similarities between CS and TS. The sentence “Some indigenous groups living in palm-rich areas use palms to make many of their necessary items and food.” from the ASSET (Alva-Manchego et al., 2020a) validation set was simplified by one of the annotators as “Groups who live in palm-rich areas use palms to make basic items and food.”. A CS, in this case, could be “People who live in areas with a lot of palm trees use the trees for many …
Content truncated to 500 words
Labeled Text
AbstractText Simplification (TS) is the task of converting a text into a form that is easier to read while maintaining the meaning of the original text. A sub-task of TS is Cognitive Simplification (CS), converting text to a form that is readily understood by people with cognitive disabilities without rendering it childish or simplistic. This sub-task has yet to be explored with neural methods in NLP, and resources for it are scarcely available. In this paper, we present a method for incorporating knowledge from the cognitive accessibility domain into a TS model, by introducing an inductive bias regarding what simplification operations to use. We show that by adding this inductive bias to a TS-trained model, it is able to adapt better to CS without ever seeing CS data, and outperform a baseline model on a traditional TS benchmark. In addition, we provide a novel test dataset for CS, and analyze the differences between CS corpora and existing TS corpora, in terms of how simplification operations are applied.
Publication Details
- Paper ID:
- 41638
- Year:
- 2022
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan