Neural Text Classification for Digital Transformation in the Financial Regulatory Domain
Authors:
- 16 November 2022
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Abstract
A core use case in artificial intelligence and natural language processing (NLP) is automatic text classification of documents, for the efficient, transparent and reliable handling of the billions of documents generated each year as part of business and government operation. Our application for document analysis uses deep learning for Neural Text Classification, with recurrent (Bi-LSTM) and transformer neural networks (DistilBERT and FinBERT). We compare the new models against traditional TF-IDF bag-of-words machine learning models, and evaluate text classification on a corpus of over 2,600,000 consumer financial complaints from the U.S. Consumer Financial Protection Bureau (CFPB), an agency of the U.S. Federal government created as a result of the 2008 financial crisis. Our analysis shows the superiority of the transformer models, with a classification accuracy of 88.05% on the task formulated.Keywords
This publication has 7 references indexed in Scilit:
- Sentiment Analysis2020
- Model Cards for Model Reporting2019
- Sentence-Level Sentiment Analysis of Financial News Using Distributed Text Representations and Multi-Instance Learning2019
- Speech and language processing2010
- Natural Language Processing for Online Applications2007
- Long Short-Term MemoryNeural Computation, 1997
- Empty Categories, Chain Binding, and Parsing1991