Deep learning and visual tools for analyzing and monitoring integrity risks

Citation

Albert Weichselbraun and Christian Hauser and Sandro Hörler and Anina Havelka. (2020). Deep learning and visual tools for analyzing and monitoring integrity risks, 5th SwissText & 16th KONVENS Joint Conference 2020

Abstract

Risks jeopardizing the integrity of an organization are widespread. According to a 2018 study by PricewaterhouseCoopers, almost 40% of Swiss companies haven been affected by illegal and unethical behavior, such as embezzlement, cybercrime, corruption, fraud, money laundering and anti-competitive agreements. Although the number of cases within Switzerland is relatively low, the financial impact of these incidents is still above the global average. The University of Applied Science of the Grisons conducts research that applies web intelligence and deep learning to the task of supporting Swiss companies in identifying and mitigating integrity risks. Historical data is used for training an LSTM classifier to recognize national and international media coverage on corruption. Afterwards, we apply transfer learning techniques to the task of adapting the classifier to a wide range of integrity topics such as human rights, labor conditions and sustainability. The adapted classifier assigns scores to News articles that indicate their relevance to the topic of integrity. Sophisticated visual tools use the annotated documents for (i) tracking and visualizing past integrity management gaps and their respective impacts, (ii) identifying whether organizations have been mentioned positively or negatively in these events, and (iii) leveraging media coverage on upcoming integrity stories for predicting and discovering existing blind spots within a company’s governance.

Downloads and Resources

  1. Reference (BibTex)
  2. Full Article