Linguistic analysis and visualization of texts of pupils to draw conclusions about higher-level characteristics
Institute(s): TU Darmstadt
External Partner: Differential Psychology & Psychological Diagnostics, Goethe University Frankfurt
Financed with funds from the project executing agency: Dec. 2014 – Nov. 2015 as pilot project
Prof. Dr. Iryna Gurevych, TU Darmstadt
Dr. Daniela Oelke, TU Darmstadt
Prof. Dr. Sonja Rohrmann, Goethe University Frankfurt
The pilot project develops an interactive, visual tool for the analysis of the relationship between linguistic properties of texts and superior features. The project is of great interest for various areas of educational research that have texts from individuals and want to make statements about the relationship between the texts and the characteristics. For example, two different populations (e.g. girls and boys, children with/without migration background), the texts of a test subject over time, or even a test subject can be compared with the average values for a certain population. Within the framework of the pilot project, an exemplary application case will be addressed in which the text analysis is linked to the personality traits of a test subject and which will later make it possible to generalise the methodology. Language can be used to express personality traits as well as cognitive, emotional and social processes. From a psychological point of view, it will be investigated in particular whether valid personality traits such as verbal and mathematical intelligence, socio-economic background and other personality traits (achievement motivation, competence and control conviction, emotional stability) can be deduced from students’ short stories. The project also provides new insights into how personality traits, socio-economic background and intelligence, among other things, are reflected in texts and to what extent these are used by third parties in order to identify characteristics of a person.
Associated Software Products
Lerner, Patrick: Designing a User Interface for Data Analysis and Feature Engineering in Text Classification. Bachelor Thesis. Technical University of Darmstadt. 2015.
Sousa, Tahir; Flekova, Lucie; Mieskes, Margot and Gurevych, Iryna: Constructive Feedback, Thinking Process and Cooperation – Assessing the Quality of Classroom Interaction. In Proceedings of INTERSPEECH Conference, p. 2739-2743. 2015.
Stoffel, Florian; Flekova, Lucie; Oelke, Daniela; Gurevych, Iryna and Keim, Daniel: Feature-Based Visual Exploration of Text Classification. In IEEE (Ed.), Proceedings of the Symposium on Visualization in Data Science (VDS) at IEEE VIS. 2015.
Hoppe, David: Write Me a Story and I Tell You Who You Are. Master Thesis. Technical University of Darmstadt. 2014.
As part of the pilot project, the so-called MINERVA tool for the visual exploration of features for text classification was further developed. The content of this tool was examined by means of transcripts to determine the success of the interaction in classrooms. For this purpose, (para-)linguistic features were extracted from the transcripts in order to train models with which the quality of teaching can be determined automatically. The developed system works with comparable precision as the experts who transcribed and annotated the recordings of the classroom situations previously.