We present results from using Probit models to classify and rank texts of varying complexity from multiple sources. We use multiple linguistic sources includingSwedish easy-to-read books and investigate data augmentation and feature regularisation as optimisation methods for text complexity assessment. Multi-Scale and Single Scale Probit models are implemented using different ratios of training data, and then compared. Overall, the findings suggest that the Multi-Scale Probit model is an effective method for classifying text complexity and ranking new texts and could be used to improve the performance on small datasets as well as normalise datasets labelled using different scales. 

With the purpose of  analysing Swedish companies' adherence and adoption of the information security standard ISO 27001 and to examine the communicative constitution of preventive innovation in organisations, we have created a corpus of corporate texts from Swedish company websites. The corpus was analysed from multiple interdisciplinary perspectives in close cooperation with management researchers and SweClarin researchers using SweClarin tools and resources as well as standard language technology tools. Some analyses require deep reading, which was performed by management researchers, often guided by results from language analyses. Initial results have been presented at a management studies conference. In this paper, we focus on presenting the research issues, the methods used in the project, the results, and the experience of SweClarin researchers supporting researchers in  social sciences. Our contribution is to show how it is possible, through the integration of human insights and digital methods, to increase the credibility and validity of a digitally acquired data set and subsequent research findings. In our view, a combination of human deep reading (management researchers), contextual lexical verification (management studies) and language technology (content and sentiment analysis) can help to sensitise computational text analysis for medium-sized data sets.

 We explore three auxiliary techniques for automatic text adaptation (ATA)—epithets for nouns, explanations for keywords, and syllabification—to aid reading for individuals with reading difficulties. In an initial evaluation, we conducted a study with individuals possessing average reading skills. Results indicate that while all three techniques demonstrate high accuracy, their usefulness varies. Epithets were found to be less beneficial, possibly due to the introduction of excessive information, although they may assist certain populations, such as individuals with intellectual disabilities. Keyword explanations were generally helpful and accurate, though occasional inaccuracies arose with rare or domain-specific terms. The effectiveness of syllabification was found to be contingent on the specific words being processed. These findings suggest that while ATA techniques can improve reading accessibility, their varying impacts highlight the need for tailored approaches based on the reader's needs.

Making information accessible to diverse target audiences, including individuals with dyslexia and cognitive disabilities, is crucial. Automatic Text Simplification (ATS) systems aim to facilitate readability and comprehension by reducing linguistic complexity. However, they often lack customizability to specific user needs, and training data for smaller languages can be scarce. This paper addresses ATS in a Swedish context, using methods that provide more control over the simplification. A dataset of Swedish paraphrases is mined from large amounts of text and used to train ATS models utilizing prefix-tuning with control prefixes. We also introduce a novel data-driven method for selecting complexity attributes for controlling the simplification and compare it with previous approaches. Evaluation of the trained models using SARI and BLEU demonstrates significant improvements over the baseline -- a fine-tuned Swedish BART model -- and compared to previous Swedish ATS results. These findings highlight the effectiveness of employing paraphrase data in conjunction with controllable generation mechanisms for simplification. Additionally, the set of explored attributes yields similar results compared to previously used attributes, indicating their ability to capture important simplification aspects.

In this paper, we present results from investigations of text complexity using cohesion measures and their importance related to other text complexity measures. To provide additional nuance, we introduce the interrelated concepts of epistemic stance and narrativity, deepening the analysis of the statistical findings. These concepts also facilitate further discussion on complexity and cohesion as they relate to reading skills and knowledge asymmetries. We employ principal component analysis (PCA) to uncover these statistical relationships on a broader scale, while conducting more specific in-depth analyses of certain metrics. Our findings, which mostly align with existing literature, reaffirm the significance of narrativity in contextualizing cohesion. However, we unexpectedly found a clear link between higher complexity and less narrative text. Additionally, the PCA reveals a more nuanced picture of referential cohesion and the use of its constituent metrics, which varies depending on both narrativity and complexity.

We investigate changes in the use of four Swedish words from the fields of design and archi- tecture. It has been suggested that their meanings have been blurred, especially in governmental reports and policy documents, so that distinctions between them that are important to stakeholders in the respective fields are lost. Specifically, we compare usage in two governmental public reports on design, one from 1999 and the other from 2015, and additionally in opinion responses to the 2015 report. Our approach is to contextualise occurrences of the words in different representations of the texts using word embeddings, topic modelling and sentiment analysis. Tools and language resources developed within the SweClarin infrastructure have been crucial for the implementation of the study.

We show how the tools and language resources developed within the SweClarin infrastructure can be used to investigate changes in the use and understanding of the Swedish related words arkitektur, design, form, and formgivning. Specifically, we compare their use in two governmental public reports on design, one from 1999 and the other from 2015. We test the hypothesis that their meaning has developed in a way that blurs distinctions that may be important to stakeholders in the respective fields.

Preventive innovation differs from ordinary innovation. The innovation diffusion literature claims that the economic benefits of preventive innovation to adopters, such as ensuring information security, are mainly intangible and often time-delayed and sometimes only adopted for incidents that may never occur. Adopter communication about preventive innovation therefore seems to be crucial.

Using the example of the information security standard ISO/IEC 27001, we examine how communication of preventive innovations is shaped by its adopters. By analyzing texts about the information security standard ISO/IEC 27001 on Swedish corporate websites using computational linguistics tools and classical content analysis, we could identify, first, different adoption approaches of preventive innovation driven, second, by three modes of data governance: agency, stewardship and brokerage. Third, we provide evidence that the communication of preventive innovation depends on its data governance mode, but, fourth, also on the economic benefits of preventive innovation for adopters.

Our contribution to the innovation literature is twofold. First, the concept of preventive innovation originally presented by Rogers (1995) is revived and further developed. Comparing it to its original scope, we show that preventive innovation can be meaningful for adopting organizations not only when they go through all possible adoption phases identified by Rogers (1995). Also an economic benefit from preventive innovation is possible. Both aspects, adoption approach as well as economic opportunity strongly shape the production of meaning in communication about preventive innovation. Second, we show that computational linguistics can support qualitative research in the study of meaning production in communication, especially when dealing with large amounts of data, for instance, gained from corporate websites.