Workflow Systems and Technology

Text Analytics for Decision Support

A large share of decision-relevant information comes in the form of unstructured text: corporate disclosures and financial news, social media posts, product reviews, real estate listings, or business documents. This research stream investigates how such data can be unlocked with methods from natural language processing (NLP) and machine learning to support better decisions.

From a predictive perspective, the research develops methods that extract actionable signals from text. Examples include trading strategies based on corporate disclosures, automated real estate valuation using property descriptions, and predictive process monitoring that incorporates digital documents as an additional source of information. From an explanatory perspective, it examines how language shapes human judgments and market outcomes. For instance, longer product reviews are not necessarily more helpful, particularly when they frequently switch between positive and negative arguments. For blockchain ventures, positive language and a high yet steady level of interaction with the community on Twitter are associated with greater funding success.

Methodologically, one focus is on fine-grained analyses. Multi-instance learning makes it possible to transfer information from the document level to the sentence level, for example to identify the tone of individual sentences in financial news or argumentation patterns in product reviews. Another emphasis is on realistic evaluation. For text-based trading strategies, for instance, profitability can be substantially overestimated if liquidity and order execution timing are not properly accounted for.

Selected publications 

  • Levich, S., Lutz, B., & Neumann, D. (2023). Utilizing the omnipresent: Incorporating digital documents into predictive process monitoring using deep neural networks. Decision Support Systems, 175, 114043. https://doi.org/10.1016/j.dss.2023.114043
  • Baur, K., Rosenfelder, M., & Lutz, B. (2023). Automated real estate valuation with machine learning models using property descriptions. Expert Systems with Applications, 213, 119147. https://doi.org/10.1016/j.eswa.2022.119147
  • Schmitz, H. C., Lutz, B., Wolff, D., & Neumann, D. (2023). When machines trade on corporate disclosures: Using text analytics for investment strategies. Decision Support Systems, 165, 113892. https://doi.org/10.1016/j.dss.2022.113892
  • Lutz, B., Pröllochs, N., & Neumann, D. (2022). Are longer reviews always more helpful? Disentangling the interplay between review length and line of argumentation. Journal of Business Research, 144, 888–901. https://doi.org/10.1016/j.jbusres.2022.02.010
  • Lutz, B., Pröllochs, N., & Neumann, D. (2020). Predicting sentence-level polarity labels of financial news using abnormal stock returns. Expert Systems with Applications, 148, 113223. https://doi.org/10.1016/j.eswa.2020.113223
  • Albrecht, S., Lutz, B., & Neumann, D. (2020). The behavior of blockchain ventures on Twitter as a determinant for funding success. Electronic Markets, 30(2), 241–257. https://doi.org/10.1007/s12525-019-00371-w
  • Pröllochs, N., Feuerriegel, S., Lutz, B., & Neumann, D. (2020). Negation scope detection for sentiment analysis: A reinforcement learning framework for replicating human interpretations. Information Sciences, 536, 205-221. https://doi.org/10.1016/j.ins.2020.05.022