Automated Financial Data Extraction from YouTube Earnings Calls

Description and use

A reproducible Python pipeline that converts a 15-minute YouTube earnings video into a structured Pandas DataFrame containing key financial metrics like Revenue, GMV, and Net Loss, enabling rapid analysis and integration into AI-driven decision-making tools.

Problem

Manually transcribing and extracting structured financial data from lengthy YouTube earnings call videos is time-consuming and prone to human error, making it difficult for developers and founders to quickly analyze competitor metrics.

Outcome

A reproducible Python pipeline that converts a 15-minute YouTube earnings video into a structured Pandas DataFrame containing key financial metrics like Revenue, GMV, and Net Loss, enabling rapid analysis and integration into AI-driven decision-making tools.

Workflow

• Install required Python packages: youtube-transcript-api, pandas, tqdm, openai, langchain.
• Fetch the YouTube video transcript using the youtube-transcript-api library and save it as a text file.
• Perform a sanity check on the transcript to ensure relevant content (e.g., fiscal year mentions) is captured.
• Define regular expression patterns for specific financial metrics (Revenue, Net Loss, GMV, Contribution Margin).
• Extract structured data from the transcript using regex and store it in a Pandas DataFrame.
• Utilize OpenAI's GPT-4 via LangChain to handle fuzzy extraction of numbers spoken in natural language that regex might miss.
• Combine and clean the extracted data into a final structured dataset for further analysis.

Evidence

We'll use the youtube-transcript-api library, which fetches the auto-generated (or creator-provided) subtitles in JSON format.

Source

We'll use a combination of regular expressions and OpenAI's GPT-4 (via LangChain) to turn those mentions into a tidy pandas DataFrame.

Source