Advanced Deep Learning Techniques for Analyzing Earnings Call Transcripts: Methodologies and Applications

TL;DR

This study investigates whether sentiment extracted from earnings-call transcripts can predict near-term stock performance by comparing multiple transformer-based NLP models, including FinBERT, BERT, ULMFiT, and Longformer. Transcripts are collected from SeekingAlpha, aligned with price movements around earnings, and labeled using a defined price-movement threshold to create a supervised task. FinBERT fine-tuning yields the strongest performance (~52% accuracy), but all models struggle with the pervasive sugar-coating in financial language, limiting their practical predictive power. The results indicate potential for NLP-driven signals in financial decision-making, yet emphasize the need for domain-specific pretraining and larger, more varied datasets to achieve reliable, real-world applicability.

Abstract

This study presents a comparative analysis of deep learning methodologies such as BERT, FinBERT and ULMFiT for sentiment analysis of earnings call transcripts. The objective is to investigate how Natural Language Processing (NLP) can be leveraged to extract sentiment from large-scale financial transcripts, thereby aiding in more informed investment decisions and risk management strategies. We examine the strengths and limitations of each model in the context of financial sentiment analysis, focusing on data preprocessing requirements, computational efficiency, and model optimization. Through rigorous experimentation, we evaluate their performance using key metrics, including accuracy, precision, recall, and F1-score. Furthermore, we discuss potential enhancements to improve the effectiveness of these models in financial text analysis, providing insights into their applicability for real-world financial decision-making.

Figures (12)

• Figure 1: Data Preparation Plots
• Figure 2: Performance Metrics for baseline FinBERT
• Figure 3: Performance Metrics for fine-tuned FinBERT
• Figure 4: Training Metrics for fine-tuned FinBERT
• Figure 5: Testing Accuracy Baseline vs Fine-Tuned FinBERT