>> AI ALLIANCE RP > AI RESEARCH > AI PROJECTS > SEDIS: SEMANTIC DISENTANGLEMENT: DISTINGUISHING STYLE FROM SUBJECT IN TEXT DATA
SeDis: Semantic Disentanglement: Distinguishing Style from Subject in Text Data
- Duration: 01.06.2020 - 31.05.2024
Funding Organization: BMBF
Jun.-Prof. Dr. Sophie Fellenz
University of Kaiserslautern-Landau
The project "Semantic Disentanglement: Distinguishing Style and Subject in Text Data" is concerned with the development of models and software to improve the automatic analysis and generation of qualitative text. Potential applications offer areas where communication between humans and machines is central, such as in customer support or social media. A specific challenge here is what exactly should be the content of the text that is generated, and controlling the style of the text separately. A "disentanglement", i.e. the untangling of style and topic in text data, should improve the influence on the generated texts and thus also on their quality.
https://press.uni-mainz.de/german-federal-ministry-of-education-and-research-is-funding-sophie-burkhardt-to-establish-a-computer-science-junior-research-group/
AI Focus Areas of the Research Project
Basic Research
- Machine Learning (ML): Self-Supervised Learning, Reinforcement Learning (RL), Unsupervised Learning
- Technology Analysis: Sociological Aspects, Human-AI Interaction