Modern automatic text adaptation often fails to meet the specific needs of targeted readers. The TextAD project aimed to address this gap by creating tailored texts for students with intellectual disability (ID), focusing on their unique challenges. This exploratory study presents findings from three rounds of testing automatic adaptations at lexical, syntactic, and discourse levels in expository texts. Reading comprehension, self-reported ratings of comprehension, interest, and perceived difficulty, along with metacognitive insights were assessed with 27 students with ID, refining adaptations based on each iteration's results. While comprehension gains for the adapted versions were not significant overall, promising outcomes appeared with lexical and syntactic adaptations in the later rounds of testing, although group heterogeneity influenced results. Despite reporting high comprehension, students noted persistent difficulty understanding many expository concepts. Alongside evaluating the effects of text adaptations on comprehension and perceived understanding, this paper offers insights into assessment approaches for developing automatic adaptations tailored to this audience.

Imagine what life would be if words did not make any sense to you. If you had to use a dictionary, or Google Translate, to check the meaning of each word. If you forgot the first part of the sentence when you got to the last word. Working would be hard, studying perhaps even more difficult. Reading is a skill essential for participation in many parts of the modern society, but there are many individuals incapable of assimilating text. This could be for various reasons, for instance due to dyslexia, cognitive disabilities, or as a consequence of having a different first language. In order to automate the adaptation process, more knowledge about the properties of Easy Language texts is needed. There are several collections of guidelines aimed at writers of Easy Language texts. Such guidelines are useful for getting an idea of what Easy Language should be, but they are often rather vague and rely on the expert knowledge and intuition of the writer. By analysing texts produced by writers of Easy Language, it is possible to operationalise the guidelines and obtain data for building models for automatic text adaptation. However, this approach requires large corpora of high quality, and the results are highly dependent on the training data.This doctoral thesis adopts an end-user approach to automatic text adaptation. The application of a reading comprehension perspective on Easy Language text adaptation differs from the text-based perspectives commonly used within the field and this position implies that the strengths and weaknesses of individuals in the different groups of poor readers are important to consider. We present and evaluate a variety of techniques for automatic text adaptation, as well as work on text complexity visualisation. Assessment of text complexity can be useful for both readers and writers of Easy Language texts. However, common complexity features are not easily interpretable, and this thesis presents work on enhancing interpretability of such features, using clustering and visualisation methods. The main contributions of this thesis are 1) a mapping of Easy Language guidelines to a theoretical model of reading comprehension, 2) a corpus of simple and standard documents, aligned at the sentence level, 3) clustering and visualisation of text complexity measures, and 4) a number of tools and services for the automatic adaptation of Swedish text.

In this paper, we present the current version of The Swedish Simplification Toolkit. The toolkit includes computational and empirical tools that have been developed along the years to explore a still neglected area of NLP, namely the simplification of “standard” texts to meet the needs of target audiences. Target audiences, such as people affected by dyslexia, aphasia, autism, but also children and second language learners, require different types of text simplification and adaptation. For example, while individual with aphasia have difficulties in reading compounds (such as arbetsmarknadsdepartement, eng. ministry of employment), second language learners struggle with cultural-specific vocabulary (e.g. konflikträdd, eng. afraid of conflicts). The toolkit allows user to selectively decide the types of simplification that meet the specific needs of the target audience they belong to. The Swedish Simplification Toolkit is one of the first attempts to overcome the one-fits-all approach that is still dominant in Automatic Text Simplification, and proposes a set of computational methods that, used individually or in combination, may help individuals reduce reading (and writing) difficulties.

In this paper, we propose visualizing results of a corpus-based study on text complexity using radar charts. We argue that the added value of this type of visualisation is the polygonal shape that provides an intuitive grasp of text complexity similarities across the registers of a corpus. The results that we visualize come from a study where we explored whether it is possible to automatically single out different facets of text complexity across the registers of a Swedish corpus. To this end, we used factor analysis as applied in Biber’s Multi-Dimensional Analysis framework. The visualization of text complexity facets with radar charts indicates that there is correspondence between linguistic similarity and similarity of shape across registers.

We present results from assessing text complexity based on a factorisation of text property measures into components. The components are evaluated by investigating their ability to classify texts belonging to different genres. Our results show that the text complexity components correctly classify texts belonging to specific genres, given that the genres adhere to certain textual conventions. We also propose a radar chart visualisation to communicate component based text complexity.

We present a method for aligning comparable corpora of simple-complex articles at the sentence level. Three methods were tested; Average Alignment (AA), Maximum Alignment (MA), and Hungarian Alignment (HA). For evaluating the algorithms, and finding the optimal combination of parameters, a dataset of manually annotated sentences was constructed. The algorithms were evaluated against the manually annotated dataset, and the best-performing algorithm proved to be the MA algorithm, which resulted in corpus comprising 59,513 aligned sentence pairs, of which 17,653 were unique sentences.

We present a language technology service for web editors’ work on making texts easier to understand, including tools for text complexity analysis, text simplification and text summarization. We also present a text analysis service focusing on measures of text complexity.

We present a rule based automatic text simplification tool for Swedish. The tool is designed to facilitate experimentation with various simplification techniques. The architecture of the tool is inspired by and partly built on a previous text simplification tool for Swedish, CogFLUX. New functionality, new operation types, and new simplification operations were added.

We present results from an eye tracking study of automatic text summarization. Automatic text summarization is a growing field due to the modern world’s Internet based society, but to automatically create perfect summaries is challenging. One problem is that extraction based summaries often have cohesion errors. By the usage of an eye tracking camera, we have studied the nature of four different types of cohesion errors occurring in extraction based summaries. A total of 23 participants read and rated four different texts and marked the most difficult areas of each text. Statistical analysis of the data revealed that absent cohesion or context and broken anaphoric reference (pronouns) caused some disturbance in reading, but that the impact is restricted to the effort to read rather than the comprehension of the text. However, erroneous anaphoric references (pronouns) were not always detected by the participants which poses a problem for automatic text summarizers. The study also revealed other potential disturbing factors.