Data Centric Approach for Multimodal Financial Data
Overview
Multimodal data is common in business, finance, accounting, and auditing.
Textual data: Text is the most prevalent data type, including financial news, financial reports, earnings conference call transcripts, and social media posts. These textual data provide timely market information and reflect market participants’ sentiments.
Numerical data: Numerical data, such as stock prices, financial indicators, and economic statistics, offer market insights. Investors and analysts frequently rely on numerical data for market forecasting.
Chart data: Charts are frequently included in financial reports, news articles, and related materials. It visually represents market trends and patterns, facilitating easier interpretation of market behavior and dynamics.
Tabular data: Structured financial data presented in tables, including balance sheets, income statements, stock prices, and trading volumes.
Time-series data: It is a sequence of data points indexed in time order. In the financial sector, time series data is commonly used to represent how a financial indicator changes over time.
Visual data: Visual data includes images and videos. They are from financial media and official announcements. Visual data provide detailed insights beyond textual and numerical data, illustrating complex market events and trends.
Audio data: Financial podcasts and recordings of earnings conference calls contain critical auditory information. Audio modalities can influence market perception and offer additional dimensions for sentiment analysis and market prediction.
Multimodal financial data can refer to a combination of the above uni-modal data. For instance, Earning Conference Calls (ECCs) consist of two modalities: the audio of a presentation and its textual transcripts. We list the common types in below table and describe them in the following subsections.
Types |
Text |
Audio |
Image |
Video |
Numbers |
Tabular |
Chart |
Time-Series |
|---|---|---|---|---|---|---|---|---|
Earnings Conference Calls (ECC) |
✓ |
✓ |
||||||
Monetary Policy Calls (MPC) |
✓ |
✓ |
✓ |
|||||
Climate Data |
✓ |
✓ |
✓ |
|||||
Financial News |
✓ |
✓ |
✓ |
✓ |
✓ |
✓ |
||
Market Data |
✓ |
✓ |
✓ |
✓ |
✓ |
|||
Financial Reports |
✓ |
✓ |
✓ |
✓ |
✓ |
✓ |
||
Financial curriculum and certificates |
✓ |
✓ |
✓ |
✓ |