Dataset

sp500_earnings_transcripts

Earnings call transcripts for S&P 500 and other large U.S. companies covering 2005 through 2025, useful for financial research, natural language processing, and sentiment analysis.

Open on Hugging Face

Data preview

Subset

Loading preview...

Dataset card

The sp500_earnings_transcripts dataset, published on Hugging Face by idleengine, is an MIT-licensed corpus of S&P 500 and US large-cap earnings call transcripts spanning 2005 to 2025. It is distributed as a single train split of 33,362 rows and is designed for financial NLP and sentiment research.

What this dataset contains

Each row corresponds to one earnings call transcript for a publicly traded company, with rows identified by ticker symbol and dated between 2005 and 2025. Coverage extends across all 11 GICS sectors and includes major constituents such as Agilent Technologies (A), Abbott Laboratories (ABT), Exxon Mobil (XOM) and Yum! Brands (YUM). The README notes that only the primary share class is typically included for companies with multiple listings.

Structure and fields

The published schema comprises eight fields: symbol and company_name as strings, quarter and year as int64, date as a string, content as the full transcript text, structured_content as a list of speaker/text segment objects enabling section-level analysis, and company_id as a float64. The structured_content field separates the operator introduction, management prepared remarks, Q&A and closing remarks, so analysts can isolate the management presentation from analyst questions.

How analysts use it

Investment teams typically use such a corpus to train or backtest NLP models that score management tone, detect guidance changes or extract forward-looking statements from the Q&A segment of structured_content. Long/short equity funds can also build factor signals by aggregating transcript-derived sentiment scores at the symbol level across the 33,362-row history, while event-driven mandates may flag transcript dates (date, quarter, year) around earnings releases to study post-call drift.

Documentation & references 1