Dataset

sp500_earnings_transcripts

A collection of earnings call transcripts covering S&P 500 firms and other U.S. large-cap companies across the years 2005 through 2025, intended for financial analysis, NLP modeling, and sentiment research.

Open on Hugging Face

Data preview

Subset

Loading preview...

Dataset card

The sp500_earnings_transcripts dataset is a Hugging Face collection of quarterly earnings call transcripts for S&P 500 and other US large-cap companies, covering roughly two decades of corporate communications. It is published under an MIT license and is primarily intended for natural language processing research on financial text.

What this dataset contains

The dataset comprises approximately 33,362 transcript records stored in a single train split, sized between 10 million and 100 million rows of text. Coverage spans 2005 through 2025 and includes constituents across all 11 GICS sectors, with the readme noting that transcript counts grow materially after 2007 and that some firms appear under their primary share class only.

Structure and fields

Each record is keyed by a string symbol and string date, with integer quarter and year fields, a string company_name, and a float company_id. The full unstructured call text is held in the string content field, while a parallel structured_content field breaks the call into segments carrying a speaker label and segment text, enabling section-aware analysis of operator remarks, prepared management remarks, and the Q&A portion.

How analysts use it

Quant funds and fundamental analysts can use the structured_content segments to track management tone and analyst questioning patterns over time, or build text-classification features aligned with stock returns. The symbol, quarter, and year fields make it straightforward to join transcripts to price, factor, or estimates data for event-style studies around earnings releases.

Documentation & references 1