Proposal: Earnings Call Analysis

Proposals Round 7
#1

The Earning Analysis project assists financial analysts with investment decision making post a company’s earnings call. Most CFOs/CEOs share important details of the company’s performance during the recent quarter in the call, including successes, challenges and growth opportunities for each business line. This information is a critical component for analysts to value the respective company and aid in their investment decision making process.

How

  • User enters a YouTube link with the call or an audio file
  • OpenAI’s Whisper model coverts the speech to text
  • FinBert-Tone (ONNX version) splits the text into sentences and generates sentiment labels across 3 tags, ‘Positive’, ‘Negative’ and Neutral
  • Sentiment Plotly graphs are generated to allow visualization of sentiment across sentences
  • FaceBook Bart model is used to give a summary of the Earnings call
  • Semantic Search is enabled using sentence transformers for the analyst to search or ask specific questions or concerns that might have been addressed in the call.

Deliverables

  • Analysis of any earnings call which can be used as part of the financial analysis toolkit.
  • Build knowledge graph to display the relationship between entities

Squad

Nick Muchi

Fixed Income and FX Trader at Vanguard

Twitter: https://twitter.com/nickmuchi

Hugging Face Space link: Earnings Call Analysis Whisperer - a Hugging Face Space by nickmuchi

4 Likes
#2

Very interesting idea. Very curious if you think the application will be able to compete with the existing tools and capture the alpha?
Also very curios if the analysis could be used to suggest what words an executive might use to improve stock’s performance.

very good write-up: How to Listen for the Hidden Data in Earnings Calls | Chicago Booth Review

Prior research suggests that “personality traits are predictive of patterns of behavior,” says Zakolyukina, citing research from University of Illinois’s Brent W. Roberts. She explains that those traits can be difficult to measure in high-profile executives: “We suggest using earnings-call transcripts to estimate CEOs’ personality trait scores.”

The researchers used their findings about executive personality traits to assess how those traits are associated with firm performance. Their findings suggest numerous connections between the characteristics of CEOs and the behaviors of the companies they lead. For example, the companies of CEOs who have extroverted personalities, as observed in the words used in the transcripts, were more likely to have lower cash flows and lower returns on assets. Top executives deemed to be more conscientious tended to run slower-growth companies, while those whose dominant personality trait was openness had companies with a greater focus on R&D.

1 Like