Detect & Describe: Deep learning of bank stress in the news

TL;DR

This paper introduces a deep learning approach that uses natural language processing to detect and describe bank distress events from news articles, providing both predictive signals and interpretability.

Contribution

It presents a novel semantic vector-based model trained on news data to predict bank distress and automatically extract relevant event descriptions, enhancing interpretability and detection accuracy.

Findings

Successfully modeled 243 distress events with 6.6 million news articles.

Enabled automatic extraction of descriptive passages for stress events.

Provided a general framework for interpreting semantic-predictive models in finance.

Abstract

News is a pertinent source of information on financial risks and stress factors, which nevertheless is challenging to harness due to the sparse and unstructured nature of natural text. We propose an approach based on distributional semantics and deep learning with neural networks to model and link text to a scarce set of bank distress events. Through unsupervised training, we learn semantic vector representations of news articles as predictors of distress events. The predictive model that we learn can signal coinciding stress with an aggregated index at bank or European level, while crucially allowing for automatic extraction of text descriptions of the events, based on passages with high stress levels. The method offers insight that models based on other types of data cannot provide, while offering a general means for interpreting this type of semantic-predictive model. We model bank distress with data on 243 events and 6.6M news articles for 101 large European banks.