Instruct-FinGPT: Financial Sentiment Analysis by Instruction Tuning of General-Purpose Large Language Models
Boyu Zhang, Hongyang Yang, Xiao-Yang Liu
TL;DR
This work tackles financial sentiment analysis by addressing numeric sensitivity and contextual understanding in LLMs. It introduces Instruct-FinGPT, an instruction-tuned LLaMA-7B model that reframes sentiment classification as a generation task and maps outputs to sentiment labels, using a small, instruction-based dataset. Across Twitter Finance and FiQA datasets, Instruct-FinGPT outperforms FinBERT and even ChatGPT, with notable gains in numerical and contextual interpretation and strong zero-shot generalization to new financial data. The approach demonstrates that limited instruction data paired with generation-based prompts can leverage general-purpose LLMs for superior, resource-efficient financial sentiment analysis, paving the way for broader, instruction-tuned financial NLP applications.
Abstract
Sentiment analysis is a vital tool for uncovering insights from financial articles, news, and social media, shaping our understanding of market movements. Despite the impressive capabilities of large language models (LLMs) in financial natural language processing (NLP), they still struggle with accurately interpreting numerical values and grasping financial context, limiting their effectiveness in predicting financial sentiment. In this paper, we introduce a simple yet effective instruction tuning approach to address these issues. By transforming a small portion of supervised financial sentiment analysis data into instruction data and fine-tuning a general-purpose LLM with this method, we achieve remarkable advancements in financial sentiment analysis. In the experiment, our approach outperforms state-of-the-art supervised sentiment analysis models, as well as widely used LLMs like ChatGPT and LLaMAs, particularly in scenarios where numerical understanding and contextual comprehension are vital.