Abschlussarbeit

Text mining of audio-visual attributes: Development of a reference input data set

In the research project “Person-focused Analysis of Architectural Design”, we study human perception of built environments. The first step is to understand which dimensions are of importance to assess and verbalize the perception. Such a vocabulary can be collected by applying experimental methods such as individual vocabulary profiling. But it could also origin from already existing literature and other text-based resources. A first text mining methodology has already been developed and a first database of audio-visual attributes already exists.

In this work, you will address the question whether text mining helps to develop a reference vocabulary which is able to generally describe audio-visual perception. More specifically, the task is to find sources of input data that provide validated and established descriptors. Those descriptors are to be compared to objective parameters from the acoustic and visual domain. Furthermore, they can be checked against subjective and individually developed attributes from experiments. To find appropriate data sources, a literature study needs to be conducted. Then, an existing text mining procedure will be customized, followed by a statistical analysis and discussion of the results.

Requirements:

- Python/Matlab skills
- High interest in statistics
- Accurate and reliable work attitude