Daria Sokova
About
My research project
LLM-based Assessment of Web Content AccessibilityAs a Leverhulme Trust Doctoral Scholar, I explore the use of large language models (LLMs) to assess how accessible web content is to specific user groups. As LLMs are capable of dealing with different types of content, they could be used to identify accessibility issues in both text and the structure of the website (e.g. code of a website), providing a single pipeline for accessibility assessment. The methodology involves analyzing the existing text and web accessibility guidelines, collaborating with accessibility compliance professionals and collecting data from websites of high societal relevance. This will inform the development of prompting and fine-tuning strategies for assessing accessibility issues. Overall, this research project aims to explore a novel method for assessing web content accessibility by leveraging the multimodal capabilities of LLMs and providing an easy-to-use tool for accessing accessibility of societally relevant websites
Supervisors
As a Leverhulme Trust Doctoral Scholar, I explore the use of large language models (LLMs) to assess how accessible web content is to specific user groups. As LLMs are capable of dealing with different types of content, they could be used to identify accessibility issues in both text and the structure of the website (e.g. code of a website), providing a single pipeline for accessibility assessment. The methodology involves analyzing the existing text and web accessibility guidelines, collaborating with accessibility compliance professionals and collecting data from websites of high societal relevance. This will inform the development of prompting and fine-tuning strategies for assessing accessibility issues. Overall, this research project aims to explore a novel method for assessing web content accessibility by leveraging the multimodal capabilities of LLMs and providing an easy-to-use tool for accessing accessibility of societally relevant websites
Publications
Digital inclusion increasingly supports adults with intellectual disabilities (ID) to participate online, yet social media posts can be difficult to understand, particularly when they contain strong emotions, slang, or non-standard writing. This paper investigates whether large language models (LLMs) can simplify social media texts to improve cognitive accessibility and preserve emotional meaning. Using an accessibility-oriented prompt based on existing guidance, posts are simplified and emotion preservation is assessed. The results suggest that many simplified posts retain the same emotions, though changes occur, especially when emotions are weakly expressed or ambiguous. Qualitative analysis shows that simplification improves fluency and structure but can also shift perceived emotion through changes to tone, formatting, and other affective cues common in social media text. The research has also revealed that different LLMs produce very different outputs.
This paper summarises the submissions of our team to the TSAR 2025 Shared Task on Readability-Controlled Text Simplification, which aims to create text simplifications that balance reduced linguistic complexity, meaning preservation, and fluency while meeting a predefined target readability level. In this work, we proposed two different methods for CEFR-controlled text simplification: a setup which employed reinforcement fine-tuning of large language models (LLMs) and a conservative lexical pipeline which relied on prompting LLMs to simplify sentences.