Jonas Gunklach, Merlin Knäble

Problem

Since the onset of the COVID-19 pandemic, data visualizations have played a vital role in informing the public by visualizing Covid case numbers and predicting future values. During this time, numerous organizations launched dashboards that provided easy access to up-to-date statistics and trends showcasting the disease’s progression. While these dashboards have flourished, they have often relied on representations that leverage visual modalities for consumption, which makes them inaccessible to people who are blind and visually impaired. In addition, organizations also rely on dashboards to communicate data insights with business users - leaving blind people out in the cold. To support blind people in the interpretation of data insights, we want to convert dashboards into text using natural language generation models.

Goals

1. Identification of the state-of-the-art literature of accessible dashboards
2. Requirements elicitation based on literature and interviews
3. Development of a system that makes dashboards / data visualizations accessible (using natural language generation) and visualizes the results appropriately
4. Evaluation of said interface with a user study. Especially qualitative methods should be applied here.

Requirements

Good programming skills in Python

Experiences in natural language processing (NLP)

Interest for natural language generation (NLG)

Fluent in English (as the thesis has to be written in English)

Contact

If you are interested in this topic and want to apply for this thesis, please contact Jonas Gunklach (jonas.gunklach@kit.edu) with a short motivation statement, your CV, and a current transcript of records. Feel free to reach out beforehand if you have any questions.