Text Simplification of College Admissions Instructions: A Professionally Simplified and Verified Corpus

Proceedings of the 29th International Conference on Computational Linguistics, pages 6505–6515 October 12–17, 2022. 6505 Text Simplification of College Admissions Instructions: A Professionally Simplified and Verified Corpus Zachary W. Taylor College of Education and Human Sciences University of Southern Mississippi z.w.taylor@usm.edu Maximus H. Chu Department of Computer Science The University of Texas at Austin maximuschu@utexas.edu Junyi Jessy Li Department of Linguistics The University of Texas at Austin jessy@utexas.edu Abstract Access to higher education is critical for minor- ity populations and emergent bilingual students. However, the language used by higher educa- tion institutions to communicate with prospec- tive students is often too complex; concretely, many institutions in the US publish admissions application instructions far above the average reading level of a typical high school graduate, often near the 13th or 14th grade level. This leads to an unnecessary barrier between stu- dents and access to higher education. This work aims to tackle this challenge via text simplification. We present PSAT (Professionally Simplified Admissions Texts), a dataset with 112 admissions instructions ran- domly selected from higher education institu- tions across the US. These texts are then profes- sionally simplified, and verified and accepted by subject-matter experts who are full-time em- ployees in admissions offices at various insti- tutions. Additionally, PSAT comes with man- ual alignments of 1,883 original-simplified sen- tence pairs. The result is a first-of-its-kind cor- pus for the evaluation and fine-tuning of text simplification systems in a high-stakes genre distinct from existing simplification resources. PSAT is available at https://doi.org/ 10.5281/zenodo.7055024. 1 Introduction Access to the higher education system for minori- tized populations, especially low-income students, first-generation in college students, students of Color, and students whose first spoken language is not English, is an important social challenge (Auerbach, 2004; Cook et al., 2012; Flores, 2010; Pérez and McDonough, 2008; Rosa, 2006; Taylor, 2018, 2020a). However, researchers have consis- tently explained that much of higher education’s communication with prospective students is too complex, too lengthy, and requires a wealth of prior knowledge of the higher education system to successfully navigate (Auerbach, 2004; Pérez and McDonough, 2008); concretely many institutions in the US publish admissions application instruc- tions at or above the 14th grade English reading comprehension level, far too high for the average prospective student or adult to read and compre- hend (Taylor, 2018, 2019, 2020b). As a result, the verbose, difficult communication places an un- necessary barrier between students and access to higher education (Ardoin, 2013; Goff et al., 2004; Hartman, 1997; Kanno, 2018; Taylor, 2017). With modern conditional text generation models, one plausible way to lower such barriers is auto- matic text simplification (Siddharthan, 2014; Alva- Manchego et al., 2020), i.e., simplify text such that it is more readable and accessible, while adhering to the texts’ original semantic content. Text simpli- fication is known to benefit a range of readers, in- cluding children (Javourey-Drevet et al., 2022) and L2 learners of English (Yano et al., 1994). Yet, no studies have explored how to simplify this informa- tion, without losing important …

Content truncated to 500 words