BERT reads sentences as ordered sequences rather than simply a jumble of words, allowing it to understand intent and excel at key language understanding tasks - such as classification, question answering, and named entity recognition.
Refinitiv Labs have been using Google's BERT model and adapting it for the nuances of financially focused language.
In this 30-min virtual lab, we will share:
- The principles on which BERT operates
- Practical advice on using BERT based on our experience
- A practical application of BERT to document classification, identifying ESG controversies in news stories
- How we think BERT could transform capital markets in the future
Complete the form now to see BERT in action!
Time: 30 Minutes
Login Details: Shared after registration
BERT in action: identifying ESG controversies in news stories
The growth in sustainable investing has lead to a surge in the demand for data on environmental, social and governance (ESG) factors. However, most of the sources from which ESG data is sourced are unstructured.
Refinitiv Labs have used the BERT deep learning approach to generate language representations for news articles, and then attempt to classify them into one of 23 different ESG controversy topics. Weighted f1 performance on a test set of about 10,000 news articles is 81%.