Extracting Financial Ratios from a 10-K with h2oGPTe
- Generative AI
- Enterprise h2oGPTe
- H2O Document AI
This lesson walks through a single example of using Extractors in Enterprise h2oGPTe: pulling specific financial ratios, including revenue growth rate, net profit margin, current ratio, return on equity, and debt to equity ratio, out of Alphabet's Form 10-K filing. It covers uploading the filing into a new Collection, then opening the document's Document AI section to define a JSON schema for each metric and choosing an LLM, in this case DeepSeek R1, to run the extraction. It closes by showing the structured JSON output under Recent Results, ready to feed into a downstream app, report, or workflow instead of manually reading the filing.
Key takeaways
- How to create a Collection and upload a document like a Form 10-K so Enterprise h2oGPTe can reference it for extraction
- What a JSON schema for an Extractor needs to specify: each metric's field name, type, description, and whether it is required
- How to choose the LLM that runs an extraction job from within the Document AI section of a document
- How to find a finished extraction's structured JSON output once the Summarize, Extract, Process job completes
- Why extracting financial ratios directly from a filing saves the manual work of reading a long report by hand
Start owning your AI