Adapting information for people with cognitive or reading dificulties is essential to foster inclusion and accessibility in society. This thesis focuses on automatic text adaptation to Easy Read (ER), including aspects of text simplification and information selection, with the goal of accelerating the work of professionals and improving access to information for vulnerable groups. Our research centres on three main aspects that are critical to achieve quality ER adaptations: resource creation, neural model adaptation and model evaluation. In this context, we address corpus alignment issues, information adaptation, as well as image and explanation integration.
Around 25% of the population in countries like Spain experiences dificulties with reading and
comprehension in their daily lives [
To address this issue, professionals work to adapt these texts into Easy Read,1 a style designed to make documents easier to understand. For Spanish, experts in the field rely on the UNE 153101:2018 EX standard, which provides guidelines on creating, adapting, and verifying Easy Read documents. Additionally, the Inclusion Europe Easy Read guidelines2 are openly available and describe the key features of an Easy Read text.
Creating or adapting a text to Easy Read involves applying various linguistic strategies to improve comprehension. Lexically, this includes avoiding dificult (e.g., “help” instead of “assist”), technical jargon or foreign words, and using quantifiers like little or much instead of large or specific numerical values. Syntactically, it involves using short sentences that convey a single idea, preferably in the second person, avoiding negations, and favouring the active voice (e.g., “You have rights. You can ask for help.”). If longer sentences are necessary, they may be broken into two lines at natural pause points. Discursively, information should be well-organized: related ideas should be grouped, complex terms explained, the text structured with clear headings that reflect the content. Additionally, the text should prioritize essential content, omitting non-critical data. The challenge here lies in determining what is considered “essential”, as this can vary depending on the needs and characteristics of the target audience.
Manually adapting content to Easy Read (ER) style is time-consuming, and the volume of daily information makes this process unfeasible with current resources. This PhD thesis aims to address this challenge by developing resources, models, and tools to support ER content generation, assisting professionals in the manual adaptation process. The research will notably explore the creation of quality
ER resources via automated alignment, ER modelling with Large Language Models (LLMs), and the evaluation of automatically adapted ER texts.
Despite the limited research in the ER field noted by [
In terms of resources, ER data in Spanish and Basque remains scarce. For Spanish, [
Automatic Text Simplification is described as a text-to-text task whose goal is to reduce the complexity of a text without altering its meaning, typically through various linguistic transformations. Early simplification systems relied on handcrafted rules developed by experts [ 16] or inferred from aligned texts [17]. These rules often involved substituting complex structures or words with simpler alternatives. As machine learning capabilities advanced, research in the field shifted toward statistical and neural models based on texts, framing the task as a monolingual machine translation problem, translating from complex to simple language, using aligned sentence pairs from parallel corpora [18].
Recent lexical simplification approaches typically involve complex word identification, substitute selection, filtering and ranking; though not all of these steps are always present in a given system. LSBert [19], for example, uses a BERT model to mask complex words and predict suitable substitutes, ranking them based on the model’s probabilities, word frequency in corpora, and cosine similarity with fastText static embeddings.
More recent methods employ controllable generation, where predefined parameters are included in the model’s input data to guide its behaviour during inference [20]. Based on this idea, this approach has been applied in models such as T5 [21], T5-Large [22], BART [20], and most recently, mT5 [23], which achieved state-of-the-art results in Spanish, English and Portuguese using a single multilingual model. Large Language Models (LLMs) such as GPT-3 have also been explored for simplification, achieving results comparable to mT5 [24].
Before the emergence of LLMs, syntactic simplification was typically approached as a sequence-tosequence task, with systems trained on parallel data derived from knowledge graphs [25], extracted from Wikipedia [26], or created by experts [27]. Due to challenges in generalizing to unseen data, DisSim [28] relies on a set of handcrafted rules. DSS [29] uses a syntactically annotated corpus to create rules, which are then applied in a translation-based system.
The ACCESS system [20] was the first to use input parameters, such as word character count, Levenshtein similarity, inverse frequency, and dependency tree depth; to guide the output of a Transformer model output [30]. Later, [31] tested the same approach to a pre-trained BART model and aligned large numbers of paraphrased sentence pairs from the CCNet corpus.
Other researchers have adopted this controlled generation method using the T5 model, incorporating additional parameters such as the sentence length ratio or dependency tree depth between input and output sentences [32, 33], with the latter achieving the best results.
From an unsupervised perspective, i.e. when training output data is unavailable, [34] propose using existing machine translation corpora, treating bridge language translations as simplified versions of the original sentences in the target language. [35] compute control parameters over noisy translations generated by pre-trained XLM models and attempt to recover the original sentence, incorporating penalties for length, concurrency, and complexity during output generation.
The emergence of Large Language Models has significantly advanced the task of text simplification by addressing one of its main challenged: the scarcity of training data. Unlike earlier models, LLMs do not require large-scale annotated corpora and can perform in an unsupervised manner.
KiS model [36] uses GPT-2 in combination with a reinforcement learning algorithm that evaluates the output based on simplicity, fluency and clarity of the response. [ 37] assess the performance of ChatGPT in comparison to other existing models, supervised and unsupervised, finding that it achieves similar results. Similarly, [38] experiment with GPT-4, testing various prompts and instructions to optimize output quality, and report performance in par with the best supervised models.
Given the high cost of human evaluation, various automatic metrics have been proposed to assess the quality of simplifications. Alva-Manchego et al. [39] recommended the use of metrics such as SARI [40], SAMSA [41], and BERTScore [42], as they show higher correlation with human judgements. Conversely, the use of traditional metrics such as BLEU [43] is discouraged, as it does not align well with human assessments in the context of text simplification. In addition to these metrics, readability formulas are commonly employed to estimate the linguistic complexity of simplified texts. The Flesch–Kincaid Grade Level (FKGL) [44] is one of the most widely used, along with its Spanish adaptation, the FernándezHuerta index [45]. While these scores ofer interpretable indicators of surface-level complexity, they have been criticized for their high sensitivity to small lexical or syntactic changes that may not significantly afect overall readability or simplification quality [ 46]. Nonetheless, they remain useful in applications aimed at producing accessible content.
This thesis aims to investigate the automatic adaptation of texts to ER, to support professional practitioners in accelerating the adaptation process and to improve access to information for vulnerable populations. The primary target languages for this research will be Basque and Spanish, with particular emphasis on Basque, where resources for text adaptation are especially scarce.
To achieve the objectives of this thesis, the development of parallel corpora (complex-adapted) is foreseen, alongside the creation of text adaptation models for various domains, including narrative, news and technical documents.
The main objectives of the thesis are: 1. Development of methods and tools for Easy Read text adaptation. While models for general text simplification exist, notable characteristics of the Easy Read style, such as sentence splitting and rephrasing, or automatic sentence segmentation, are relatively less explored in end-to-end fashion. This thesis will explore methods and tools specifically designed to cover these aspects. 2. Creation of an automatic complex-adapted text aligner. Both Spanish and, more notably, Basque, currently lack parallel corpora for automatic text adaptation. This scarcity hinders progress in the field. An objective of this thesis is the development of components and methods to align ER texts at diferent levels of granularity (e.g., sentence and paragraph levels), to provide quality ER-adapted alignments and enable the creation of new corpora. 3. Development of ER datasets in Spanish and Basque across domains. Using the aforementioned aligner, new training datasets will be created for a variety of domains. This is essential, as the Easy Read adaptation style can vary between domains. 4. Training and benchmarking of adapted neural models. New neural models will be trained using both the corpora developed in this thesis and publicly available data. These models will be benchmarked against existing approaches in the literature to assess their efectiveness. 5. Development of methods and models for the automatic generation of domain-adapted explanations of complex terms. Easy Read texts frequently include brief explanations of complex terms, particularly when no simpler synonyms are available. This thesis explores methods and models that can identify complex terms and generate appropriate explanations in the Easy Read style, adapted to the context and domain of the text, as well as the target audience. 6. Development of methods and models for the automatic insertion of relevant, illustrative images. Easy Read texts are often supplemented with images that reinforce textual content and aid comprehension. Methods and models will be developed to automatically retrieve and assign suitable images to accompany the text, streamlining the adaptation process for professionals. Based on these objectives, the following hypotheses will be tested: 1. The development of resources, methods and models tailored to Easy Read adaptation will improve the readability of texts for people with disabilities. 2. The creation of an automatic aligner tailored to ER text will improve the quality of aligned data and, consequently, the outputs of the models trained on this data. 3. Increasing the amount of training data will enhance the quality and consistency of text adaptations. 4. The inclusion of explanations for complex terms will improve the comprehensibility of adapted texts. 5. Automatically inserting relevant images will enhance the reading experience for individuals with reading dificulties. 6. The use of neural models, particularly large language models, will support the efective generation of Easy Read content.
For model training and benchmarking, various approaches will be compared, including instruction-tuned models (few-shot and zero-shot), and fine-tuned base models. These will be evaluated against existing state-of-the-art systems across multiple datasets in Spanish and Basque, where available. Evaluation will be conducted using established metrics for text adaptation, such as SARI, BLEU, and BERTScore. Given the active research in this field, additional metrics may be incorporated if they contribute meaningfully to the interpretability and robustness of the results.
To develop the automatic complex–adapted text aligner, we will contrast various methods, including embedding and lexical similarity, as well as models tuned to the specifics of ER alignment. Sequentially, regarding the training of explanation generation models, the use of task-specific architectures or the integration of explanation generation into the main simplification and adaptation models will be investigated. These models will be designed to identify complex terms and generate suitable explanations either from a structured knowledge base or based on the surrounding context. As there are, to the best of the author’s knowledge, no publicly available datasets for this task, initial experimentation will focus on few-shot and zero-shot paradigms. Dataset creation will be considered if no suitable resource is found. Following this, an image generation module will be developed to enhance Easy Read texts. This module will either use a text-to-image generation model or retrieve relevant images from a curated image bank. Both approaches will be explored to assess their feasibility and impact on comprehension.
This research will follow standard scientific procedures, including systematic reviews of state-of-theart literature and the dissemination of findings through high-quality journals, conferences, congresses, and workshops, thereby ensuring validation by the scientific community.
Aligned with the objectives of the thesis, three topics have already been explored: 1. Split and Rephrase. One of the subtasks involved in syntactic and end-to-end text simplification and adaptation is Split and Rephrase, which takes a single complex sentence as input and breaks it into multiple shorter sentences while preserving the original meaning and information. We explored the capabilities of large language models for this task, demonstrating improvements over the state of the art. The study involved comparisons across diferent prompting strategies, domain adaptation settings, base model sizes, training data configurations, and few-shot/zero-shot approaches using instruction-tuned models. Full details can be found in [47]. 2. Text Segmentation. Easy Read guidelines specify that each sentence should be presented on its own line. Furthermore, sentemces should be split “where people would pause when reading out loud”, as recommended by the Inclusion Europe English guidelines. We introduced a novel task: automatic segmentation of sentences to conform to Easy Read principles. Two approaches were evaluated: (1) scoring-based segmentation using constituency parsers and Masked Language Model (MLM) scoring methods, and (2) generative LLM-based segmentation using in-context learning and fine-tuning. Further details are available in [48]. 3. Corpus creation. Support for ER text generation remains limited, with few resources available for the development of automated systems. We created a novel corpus of ER news texts in Basque and Spanish, marking the first publicly accessible resource designed to support the training and evaluation of ER text adaptation models in both languages. The work outlines the methodology used to build the corpus and includes both intrinsic and extrinsic evaluations. The resource is intended to be released to the research community under a CC-BY-NC-ND 4.0 license. As of this writing, the associated publication is under review.
As previously presented, this thesis aims to explore automatic text adaptation in Basque and Spanish, following Easy Read guidelines. While several experiments have been carried out, some key research questions and unresolved challenges remain: • What are the optimal criteria to identify information that needs to be preserved, omitted or transformed in ER adaptation? • How can the quality of complex-adapted alignment be evaluated accurately? • What is the best strategy to generate domain-adapted explanations for complex terms? • How can the hallucination problem be detected and mitigated in the generation of adapted or explanatory content? • How can image generation or selection be guided in a way that supports text comprehension without introducing ambiguity or distraction? • To what extent do existing evaluation metrics reflect the actual accessibility improvements brought by automated methods?
This work was partially supported by the Department of Economic Development and Competitiveness of the Basque Government (Spri Group) via funding for the IRAZ project (ZL-2024/00570).
The author has not employed any Generative AI tools. of the Workshop on Text Simplification, Accessibility, and Readability (TSAR-2022), Association for Computational Linguistics, Abu Dhabi, United Arab Emirates (Virtual), 2022, pp. 86–97. URL: https://aclanthology.org/2022.tsar-1.8/. doi:10.18653/v1/2022.tsar-1.8. [15] W. Xu, C. Callison-Burch, C. Napoles, Problems in Current Text Simplification Research: New Data Can Help, Transactions of the Association for Computational Linguistics 3 (2015) 283–297. URL: https://doi.org/10.1162/tacl_a_00139. doi:10.1162/tacl_a_00139. arXiv:https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00139/1566780/tacl_ [16] R. Chandrasekar, C. Doran, B. Srinivas, Motivations and Methods for Text Simplification, in: COLING 1996 Volume 2: The 16th International Conference on Computational Linguistics, 1996.
URL: https://aclanthology.org/C96-2183/. [17] R. Chandrasekar, B. Srinivas, Automatic induction of rules for text simplification, KnowledgeBased Systems 10 (1997) 183–190. URL: https://www.sciencedirect.com/science/article/pii/ S0950705197000294. doi:https://doi.org/10.1016/S0950-7051(97)00029-4. [18] L. Specia, Translating from Complex to Simplified Sentences, in: T. A. S. Pardo, A. Branco, A. Klautau, R. Vieira, V. L. S. de Lima (Eds.), Computational Processing of the Portuguese Language, Springer Berlin Heidelberg, Berlin, Heidelberg, 2010, pp. 30–39. [19] J. Qiang, Y. Li, Y. Zhu, Y. Yuan, Y. Shi, X. Wu, LSBert: Lexical Simplification Based on BERT, IEEE/ACM Transactions on Audio, Speech, and Language Processing 29 (2021) 3064–3076. doi:10. 1109/TASLP.2021.3111589. [20] L. Martin, É. de la Clergerie, B. Sagot, A. Bordes, Controllable Sentence Simplification, in: N. Calzolari, F. Béchet, P. Blache, K. Choukri, C. Cieri, T. Declerck, S. Goggi, H. Isahara, B. Maegaard, J. Mariani, H. Mazo, A. Moreno, J. Odijk, S. Piperidis (Eds.), Proceedings of the Twelfth Language Resources and Evaluation Conference, European Language Resources Association, Marseille, France, 2020, pp. 4689–4698. URL: https://aclanthology.org/2020.lrec-1.577/. [21] K. C. Sheang, D. Ferrés, H. Saggion, Controllable Lexical Simplification for English, in: S. Štajner, H. Saggion, D. Ferrés, M. Shardlow, K. C. Sheang, K. North, M. Zampieri, W. Xu (Eds.), Proceedings of the Workshop on Text Simplification, Accessibility, and Readability (TSAR-2022), Association for Computational Linguistics, Abu Dhabi, United Arab Emirates (Virtual), 2022, pp. 199–206. URL: https://aclanthology.org/2022.tsar-1.19/. doi:10.18653/v1/2022.tsar-1.19. [22] M. Maddela, F. Alva-Manchego, W. Xu, Controllable Text Simplification with Explicit Paraphrasing, in: K. Toutanova, A. Rumshisky, L. Zettlemoyer, D. Hakkani-Tur, I. Beltagy, S. Bethard, R. Cotterell, T. Chakraborty, Y. Zhou (Eds.), Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Association for Computational Linguistics, Online, 2021, pp. 3536–3553. URL: https://aclanthology.org/2021. naacl-main.277/. doi:10.18653/v1/2021.naacl-main.277. [23] K. C. Sheang, H. Saggion, Multilingual Controllable Transformer-Based Lexical Simplification,
Procesamiento del Lenguaje Natural 71 (2023) 109–123. [24] D. Aumiller, M. Gertz, UniHD at TSAR-2022 Shared Task: Is Compute All We Need for Lexical Simplification?, in: S. Štajner, H. Saggion, D. Ferrés, M. Shardlow, K. C. Sheang, K. North, M. Zampieri, W. Xu (Eds.), Proceedings of the Workshop on Text Simplification, Accessibility, and Readability (TSAR-2022), Association for Computational Linguistics, Abu Dhabi, United Arab Emirates (Virtual), 2022, pp. 251–258. URL: https://aclanthology.org/2022.tsar-1.28/. doi:10. 18653/v1/2022.tsar-1.28. [25] S. Narayan, C. Gardent, S. B. Cohen, A. Shimorina, Split and Rephrase, in: M. Palmer, R. Hwa, S. Riedel (Eds.), Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, Association for Computational Linguistics, Copenhagen, Denmark, 2017, pp. 606–616.
URL: https://aclanthology.org/D17-1064/. doi:10.18653/v1/D17-1064. [26] J. A. Botha, M. Faruqui, J. Alex, J. Baldridge, D. Das, Learning to Split and Rephrase From Wikipedia Edit History, in: E. Rilof, D. Chiang, J. Hockenmaier, J. Tsujii (Eds.), Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, Association for Computational Linguistics, Brussels, Belgium, 2018, pp. 732–737. URL: https://aclanthology.org/D18-1080/. doi:10. 18653/v1/D18-1080. [27] Y. Gao, T.-H. Huang, R. J. Passonneau, ABCD: A Graph Framework to Convert Complex Sentences to a Covering Set of Simple Sentences, in: C. Zong, F. Xia, W. Li, R. Navigli (Eds.), Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), Association for Computational Linguistics, Online, 2021, pp. 3919–3931. URL: https://aclanthology.org/2021.acl-long.303/. doi:10.18653/v1/2021.acl-long.303. [28] C. Niklaus, M. Cetto, A. Freitas, S. Handschuh, DisSim: A Discourse-Aware Syntactic Text Simplification Framework for English and German, in: K. van Deemter, C. Lin, H. Takamura (Eds.), Proceedings of the 12th International Conference on Natural Language Generation, Association for Computational Linguistics, Tokyo, Japan, 2019, pp. 504–507. URL: https://aclanthology.org/ W19-8662/. doi:10.18653/v1/W19-8662. [29] E. Sulem, O. Abend, A. Rappoport, Simple and Efective Text Simplification Using Semantic and Neural Methods, in: I. Gurevych, Y. Miyao (Eds.), Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Association for Computational Linguistics, Melbourne, Australia, 2018, pp. 162–173. URL: https://aclanthology.org/P18-1016/. doi:10.18653/v1/P18-1016. [30] A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. u. Kaiser, I. Polosukhin, Attention is All you Need, in: I. Guyon, U. V. Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, R. Garnett (Eds.), Advances in Neural Information Processing Systems, volume 30, Curran Associates, Inc., 2017. URL: https://proceedings.neurips.cc/paper_files/paper/2017/file/ 3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf. [31] L. Martin, A. Fan, É. de la Clergerie, A. Bordes, B. Sagot, MUSS: Multilingual Unsupervised Sentence Simplification by Mining Paraphrases, in: N. Calzolari, F. Béchet, P. Blache, K. Choukri, C. Cieri, T. Declerck, S. Goggi, H. Isahara, B. Maegaard, J. Mariani, H. Mazo, J. Odijk, S. Piperidis (Eds.), Proceedings of the Thirteenth Language Resources and Evaluation Conference, European Language Resources Association, Marseille, France, 2022, pp. 1651–1664. URL: https://aclanthology. org/2022.lrec-1.176/. [32] K. C. Sheang, H. Saggion, Controllable Sentence Simplification with a Unified Text-to-Text Transfer Transformer, in: A. Belz, A. Fan, E. Reiter, Y. Sripada (Eds.), Proceedings of the 14th International Conference on Natural Language Generation, Association for Computational Linguistics, Aberdeen, Scotland, UK, 2021, pp. 341–352. URL: https://aclanthology.org/2021.inlg-1.38/. doi:10.18653/ v1/2021.inlg-1.38. [33] S. Štajner, D. Ferrés, M. Shardlow, K. North, M. Zampieri, H. Saggion, Lexical simplification benchmarks for English, Portuguese, and Spanish, Frontiers in Artificial Intelligence Volume 5 2022 (2022). URL: https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai. 2022.991242. doi:10.3389/frai.2022.991242. [34] X. Lu, J. Qiang, Y. Li, Y. Yuan, Y. Zhu, An Unsupervised Method for Building Sentence Simplification Corpora in Multiple Languages, in: M.-F. Moens, X. Huang, L. Specia, S. W.-t. Yih (Eds.), Findings of the Association for Computational Linguistics: EMNLP 2021, Association for Computational Linguistics, Punta Cana, Dominican Republic, 2021, pp. 227–237. URL: https://aclanthology.org/ 2021.findings-emnlp.22/. doi: 10.18653/v1/2021.findings-emnlp.22. [35] O. Kariuk, D. Karamshuk, CUT: Controllable Unsupervised Text Simplification, CoRR abs/2012.01936 (2020). URL: https://arxiv.org/abs/2012.01936. arXiv:2012.01936. [36] P. Laban, T. Schnabel, P. Bennett, M. A. Hearst, Keep it Simple: Unsupervised Simplification of MultiParagraph Text, in: C. Zong, F. Xia, W. Li, R. Navigli (Eds.), Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), Association for Computational Linguistics, Online, 2021, pp. 6365–6378. URL: https://aclanthology.org/2021.acl-long.498/. doi:10.18653/v1/ 2021.acl-long.498. [37] S. Agrawal, M. Carpuat, Controlling Pre-trained Language Models for Grade-Specific Text Simplification, in: H. Bouamor, J. Pino, K. Bali (Eds.), Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, Association for Computational Linguistics, Singapore, 2023, pp. 12807–12819. URL: https://aclanthology.org/2023.emnlp-main.790/. doi:10.18653/v1/2023.emnlp-main.790. [38] X. Wu, Y. Arase, An In-depth Evaluation of Large Language Models in Sentence Simpliifcation with Error-based Human Assessment, 2025. URL: https://arxiv.org/abs/2403.04963. arXiv:2403.04963. [39] F. Alva-Manchego, C. Scarton, L. Specia, The (Un)Suitability of Automatic Evaluation Metrics for Text Simplification, Computational Linguistics 47 (2021) 861–889. URL: https://aclanthology.org/ 2021.cl-4.28/. doi:10.1162/coli_a_00418. [40] W. Xu, C. Napoles, E. Pavlick, Q. Chen, C. Callison-Burch, Optimizing Statistical Machine Translation for Text Simplification, Transactions of the Association for Computational Linguistics 4 (2016) 401–415. URL: https://aclanthology.org/Q16-1029/. doi:10.1162/tacl_a_00107. [41] E. Sulem, O. Abend, A. Rappoport, Semantic structural evaluation for text simplification, in: M. Walker, H. Ji, A. Stent (Eds.), Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers), Association for Computational Linguistics, New Orleans, Louisiana, 2018, pp. 685–696.
URL: https://aclanthology.org/N18-1063/. doi:10.18653/v1/N18-1063. [42] T. Zhang*, V. Kishore*, F. Wu*, K. Q. Weinberger, Y. Artzi, BERTScore: Evaluating Text Generation with BERT, in: International Conference on Learning Representations, 2020. URL: https://openreview.net/forum?id=SkeHuCVFDr. [43] K. Papineni, S. Roukos, T. Ward, W.-J. Zhu, Bleu: a method for automatic evaluation of machine translation, in: P. Isabelle, E. Charniak, D. Lin (Eds.), Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, Association for Computational Linguistics, Philadelphia, Pennsylvania, USA, 2002, pp. 311–318. URL: https://aclanthology.org/P02-1040/. doi:10.3115/1073083.1073135. [44] R. Flesch, A new readability yardstick., Journal of applied psychology 32 (1948) 221. [45] J. Fernández Huerta, Medidas sencillas de lecturabilidad, Consigna 214 (1959) 29–32. [46] T. Tanprasert, D. Kauchak, Flesch-Kincaid is Not a Text Simplification Evaluation Metric, in: A. Bosselut, E. Durmus, V. P. Gangal, S. Gehrmann, Y. Jernite, L. Perez-Beltrachini, S. Shaikh, W. Xu (Eds.), Proceedings of the 1st Workshop on Natural Language Generation, Evaluation, and Metrics (GEM 2021), Association for Computational Linguistics, Online, 2021, pp. 1–14. URL: https://aclanthology.org/2021.gem-1.1/. doi:10.18653/v1/2021.gem-1.1. [47] D. Ponce, T. Etchegoyhen, J. Calleja, H. Gete, Split and Rephrase with Large Language Models, in: L.-W. Ku, A. Martins, V. Srikumar (Eds.), Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Association for Computational Linguistics, Bangkok, Thailand, 2024, pp. 11588–11607. URL: https://aclanthology.org/2024.acl-long. 622/. doi:10.18653/v1/2024.acl-long.622. [48] J. Calleja, T. Etchegoyhen, D. Ponce, Automating Easy Read Text Segmentation, in: Y. Al-Onaizan, M. Bansal, Y.-N. Chen (Eds.), Findings of the Association for Computational Linguistics: EMNLP 2024, Association for Computational Linguistics, Miami, Florida, USA, 2024, pp. 11876–11894. URL: https://aclanthology.org/2024.findings-emnlp.694/. doi: 10.18653/v1/2024.findings-emnlp. 694.