Natural Language Processing in Finance: Applications and Opportunities.
Access status:
Open Access
Type
ThesisThesis type
Doctor of PhilosophyAuthor/s
Lee, JeanAbstract
The research of Natural Language Processing(NLP) in Finance has experienced considerable development driven by academia and industry. However, small benchmark datasets in financial NLP often yield lower performance in real-world scenarios, and data often requires multimodal ...
See moreThe research of Natural Language Processing(NLP) in Finance has experienced considerable development driven by academia and industry. However, small benchmark datasets in financial NLP often yield lower performance in real-world scenarios, and data often requires multimodal understanding. Moreover, financial NLP research typically focuses on single tasks, whereas real-world applications demand the integration of multiple tasks. To address these challenges, this research explores various financial NLP applications through three distinct segments. The first segment focuses on an interpretable multi-component NLP system, aiming to identify how NLP components can assist end-users in analyzing financial documents and news. This system integrates various NLP components, including sentiment analysis, topic modelling, prediction, explanation, and summarization. The second segment focuses on a novel emotion dataset in the stock market, extending beyond financial sentiment classification and demonstrating the importance of emotions in market behaviour. The dataset's usability is shown through empirical analysis, sentiment/emotion classification, and its potential for market forecasting. The third segment focuses on a multimodal financial document understanding and document QA system. This complex system integrates several deep learning models to address tasks, such as intent classification and slot filling, layout analysis, key information extraction, and Retrieval Augmented Generation(RAG) using Large Language Models(LLMs). This approach enables comprehensive financial document analysis that combines text, images, and tables. Additionally, this research provides a comprehensive review of LLMs in Finance(FinLLMs) and discusses opportunities and challenges in practical applications. By addressing the challenges of interpretability, multimodal data, and document understanding, this research aims to enhance financial decision-making processes through advanced NLP.
See less
See moreThe research of Natural Language Processing(NLP) in Finance has experienced considerable development driven by academia and industry. However, small benchmark datasets in financial NLP often yield lower performance in real-world scenarios, and data often requires multimodal understanding. Moreover, financial NLP research typically focuses on single tasks, whereas real-world applications demand the integration of multiple tasks. To address these challenges, this research explores various financial NLP applications through three distinct segments. The first segment focuses on an interpretable multi-component NLP system, aiming to identify how NLP components can assist end-users in analyzing financial documents and news. This system integrates various NLP components, including sentiment analysis, topic modelling, prediction, explanation, and summarization. The second segment focuses on a novel emotion dataset in the stock market, extending beyond financial sentiment classification and demonstrating the importance of emotions in market behaviour. The dataset's usability is shown through empirical analysis, sentiment/emotion classification, and its potential for market forecasting. The third segment focuses on a multimodal financial document understanding and document QA system. This complex system integrates several deep learning models to address tasks, such as intent classification and slot filling, layout analysis, key information extraction, and Retrieval Augmented Generation(RAG) using Large Language Models(LLMs). This approach enables comprehensive financial document analysis that combines text, images, and tables. Additionally, this research provides a comprehensive review of LLMs in Finance(FinLLMs) and discusses opportunities and challenges in practical applications. By addressing the challenges of interpretability, multimodal data, and document understanding, this research aims to enhance financial decision-making processes through advanced NLP.
See less
Date
2024Rights statement
The author retains copyright of this thesis. It may only be used for the purposes of research and study. It must not be used for any other purposes and may not be transmitted or shared with others without prior permission.Faculty/School
Faculty of Engineering, School of Computer ScienceAwarding institution
The University of SydneyShare