Can machine beat human in detecting corporate crises? Being able to detect adverse events discussed online and alerting companies in time are crucial for crisis management. During this talk, I will discuss a crisis predict and alert system FTI Consulting developed for one of the largest global e-commerce platforms by using the semantic and deep learning methods.
Being able to detect adverse events discussed online and alerting companies in the earliest time are crucial for crisis management and we have thus implemented diverse data science approaches for capturing adverse events, quantifying crisis severity and sending alerts to clients. FTI Consulting’s Data Science team has developed a data-driven solution to quantify how “bad” current business situation looks like and suggests our clients if they need take action on specific adverse events.
This talk has the following sections:
* Capture adverse event. We employed an ontology-based method to define the scope of crisis events. Following the ontological framework, a deep-learning approach was adopted to populate the ontologies with words, concepts and relations.
* Quantify crisis severity. We trained the crisis scorers using BERT - the deep learning based NLP model.
Dr. Szu-Yao Chien is a Senior Data Scientist at FTI Consulting in the UK. She possesses advanced knowledge of machine learning and applied statistics. Szu-Yao leads the technical solutions of Communications-Analytics and Crisis Alert and Predict Systems. She is engaged in FTI’s digital transformation by developing a web service solution which automates the processes of data analysis and business report generation. Before that, she worked as an NLP Data Scientist at a startup based in Manchester and was a full-time researcher in the area of Business Analytics and Data Science at the University of Manchester, where she completed her Ph.D.