Fine-Tuning Large Language Models to Improve Financial Sentiment Analysis
A State-of-the-Art Financial Sentiment Model for Enhanced Market Decision-Making
Core Research Question: Could "Generalist Geniuses" outperform highly trained "Domain Specialists" in financial sentiment analysis?
1. Research objectives
- Multi-Model Comparative Framework
- Parameter-Efficient Fine-tuning in Financial Sentiment Analysis
- Closed-Loop Evaluation Framework
- Error & Domain Adaptation Analysis
2. Major Milestone (Completed)
- Comprehensive Literature Review ✅
- Technology Choice Finalization ✅
- Set up Experiment Environment ✅
- Datasets Acquisition & Preprocessing ✅
- Traditional Benchmark Evaluation (VADER) ✅
- Domain Specialist Benchmark Evaluation (FinBERT-SA) ✅
- Parameter-Efficient Fine-Tuning (Mistral-7B & Llama3-8B) ✅
- Hyperparameter Tuning ✅
- Error Analysis ✅
- Market Validation ✅