A0221
Title: Real estate pricing and natural language processing: Improving price valuation using textual descriptions
Authors: Katharina Baur - Albert-Ludwigs-Universität (Germany) [presenting]
Abstract: A text mining approach is suggested for the valuation of residential properties. Investment decisions in the real estate sectors are generally of long-term nature. Accurate and fair pricing is, therefore, fundamental to prevent miscalculations of such investments. As the focus of previous research lies on price predictions with solely numerical features, a new approach is addressed by including written descriptions about real estate properties for assessing its value. Essentially, we make use of state-of-the-art natural language processing methods, namely contextualized embeddings, to employ textual information. We show that prediction models can significantly benefit from these methods compared to standard approaches.
