Enhancing Financial Sentiment Analysis via Retrieval Augmented Large Language Models
Boyu Zhang1*, Hongyang (Bruce) Yang2*, Tianyu Zhou3*, Ali Babar1, Xiao-Yang Liu2§, 1The University of Adelaide 2Columbia University 3Brown University Email:
Abstract
Financial sentiment analysis is critical for valuation and investment decision-making. Traditional NLP models, however, are limited by their parameter size and the scope of their training datasets, which hampers their generalization capabilities and effectiveness in this field. Recently, Large Language Models (LLMs) pre-trained on extensive corpora have demonstrated superior performance across various NLP tasks due to their commendable zero-shot abilities. Yet, directly applying LLMs to financial sentiment analysis presents challenges: The discrepancy between the pre-training objective of LLMs and predicting the sentiment label can compromise their predictive performance. Furthermore, the succinct nature of financial news, often devoid of sufficient context, can significantly diminish the reliability of LLMs’ sentiment analysis. To address these challenges, we introduce a retrieval-augmented LLMs framework for financial sentiment analysis. This framework includes an instruction-tuned LLMs module, which ensures LLMs behave as predictors of sentiment labels, and a retrieval-augmentation module which retrieves additional context from reliable external sources. Benchmarked against trad
中文速览
金融新闻和推文往往极为简短、缺乏背景，而现有大语言模型（LLM）的预训练目标又与"输出情感标签"的任务存在错位，导致直接用 ChatGPT、LLaMA 等模型做金融情感分析（financial sentiment analysis）效果不稳定。针对这两个痛点，研究者提出了一套"检索增强大语言模型"框架：先用少量专门设计的指令样本对开源 LLM 进行指令微调（instruction tuning），让模型能可靠地输出正面、负面、中性三类标签；再通过检索增强生成（RAG）模块，从彭博、路透等权威金融来源实时抓取相关背景信息，补充进输入提示，为模型提供足够的上下文。在多个金融情感分析基准上与传统小模型及 ChatGPT、LLaMA 的对比实验表明，该方法在准确率和 F1 分数上取得了 15%–48% 的显著提升，为将大语言模型落地金融智能分析提供了一条切实可行的路径。
原文 arXiv:2310.04027;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2310.04027v2