Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ECC Analyzer: Extract Trading Signal from Earnings Conference Calls using Large Language Model for Stock Performance Prediction

Yupeng Cao, Zhi Chen, Qingyun Pei, Nathan Jinseok Lee, K. P. Subbalakshmi, Papa Momar Ndiaye

TL;DR

This research introduces a novel framework: ECC Analyzer, which utilizes large language models (LLMs) to extract richer, more predictive content from ECCs to aid the model’s prediction performance and demonstrates that the model outperforms traditional analytical benchmarks, confirming the effectiveness of advanced LLM techniques in financial analysis.

Abstract

In the realm of financial analytics, leveraging unstructured data, such as earnings conference calls (ECCs), to forecast stock volatility is a critical challenge that has attracted both academics and investors. While previous studies have used multimodal deep learning-based models to obtain a general view of ECCs for volatility predicting, they often fail to capture detailed, complex information. Our research introduces a novel framework: \textbf{ECC Analyzer}, which utilizes large language models (LLMs) to extract richer, more predictive content from ECCs to aid the model's prediction performance. We use the pre-trained large models to extract textual and audio features from ECCs and implement a hierarchical information extraction strategy to extract more fine-grained information. This strategy first extracts paragraph-level general information by summarizing the text and then extracts fine-grained focus sentences using Retrieval-Augmented Generation (RAG). These features are then fused through multimodal feature fusion to perform volatility prediction. Experimental results demonstrate that our model outperforms traditional analytical benchmarks, confirming the effectiveness of advanced LLM techniques in financial analysis.

ECC Analyzer: Extract Trading Signal from Earnings Conference Calls using Large Language Model for Stock Performance Prediction

TL;DR

This research introduces a novel framework: ECC Analyzer, which utilizes large language models (LLMs) to extract richer, more predictive content from ECCs to aid the model’s prediction performance and demonstrates that the model outperforms traditional analytical benchmarks, confirming the effectiveness of advanced LLM techniques in financial analysis.

Abstract

In the realm of financial analytics, leveraging unstructured data, such as earnings conference calls (ECCs), to forecast stock volatility is a critical challenge that has attracted both academics and investors. While previous studies have used multimodal deep learning-based models to obtain a general view of ECCs for volatility predicting, they often fail to capture detailed, complex information. Our research introduces a novel framework: \textbf{ECC Analyzer}, which utilizes large language models (LLMs) to extract richer, more predictive content from ECCs to aid the model's prediction performance. We use the pre-trained large models to extract textual and audio features from ECCs and implement a hierarchical information extraction strategy to extract more fine-grained information. This strategy first extracts paragraph-level general information by summarizing the text and then extracts fine-grained focus sentences using Retrieval-Augmented Generation (RAG). These features are then fused through multimodal feature fusion to perform volatility prediction. Experimental results demonstrate that our model outperforms traditional analytical benchmarks, confirming the effectiveness of advanced LLM techniques in financial analysis.
Paper Structure (25 sections, 10 equations, 2 figures, 3 tables)

This paper contains 25 sections, 10 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Stock Volatility Prediction
  4. Large Language Models in Financial Application
  5. Problem Formulation
  6. Our proposed framework
  7. Audio Encoding
  8. Transcript Encoding
  9. Fine-grained Information Extraction from the Transcript by Using LLM
  10. Summarize the transcript
  11. Important sentences extracted by using RAG
  12. Multimodal Fusion and Prediction
  13. Experiment
  14. Dataset
  15. Baseline Setup
  16. ...and 10 more sections

Figures (2)

  • Figure 1: illustrates the ECC Analyzer Framework. The proposed method accepts multimodal inputs: audio record and transcript. The upper part of the box illustrates the feature extraction process for both the audio and text of the data. We use the pre-trained large models to generate embeddings from the audio and text, followed by a transformer encoder to extract the corresponding features. The lower part of the box represents a deeper analysis of the ECC. Here, the text is divided into chunks, which are then summarized into paragraphs using an LLM. Key sentences are extracted via RAG and converted into text features through text embedding. These text features are then fused with the features extracted from the upper part of the box to make the final volatility prediction.
  • Figure 2: visualizes the process of fine-grained information extraction from ECC transcript.