Name of the participant: Laura Seiffe
Description of the IT research project:The medium text is central when it comes to conveying information – but for the information to be conveyed successfully, the text must also be understood. In addition to other influencing factors, text complexity is particularly relevant for text comprehension. This is especially true when a technical language is used, for example in the tax domain served by DATEV. In this case, a technical text may be understandable for experts, but too complex for laypersons in this field.
Text complexity is mainly expressed on three linguistic levels: syntax, lexicon and discourse structure. Variations on these levels cause shifts in the complexity level of a text. A text can only ever be evaluated in relation to the recipient and the text domain in its complexity.
If the recipient group is heterogeneous in terms of expertise, it must be ensured that the information expressed by the text reaches everyone. Since it can be assumed that laypersons can handle a lower complexity of the text than experts, a differentiation in the complexity level of the text is helpful.
The aim of the AuTexx project is to develop models in an iterative, user-centered development process that evaluate the complexity of a tax law text with regard to an expert and a lay target group. These models are based on corresponding data sets and are the result of machine learning processes. An application is to be provided that gives the editor prompt feedback on the complexity of his text. The manual revision of the text is thus supported and simplified. Such an application reduces redundant work steps, saves resources and aligns the text with its target group in the best possible way.
Software Campus partners: DFKI, DATEV
Implementation period: 1.3.2020 – 31.12.2021