Helios Comprehend - Voice Sentiment Data (USA) - Emotion Signals Gathered from Earnings Calls

Helios Comprehend provides quantitative data that helps quants improve their trading strategies, and qualitative summaries that help fundamental researchers improve their insights into specific equities. Helios comprehend extracts 104 characteristics from an executive’s voice (e.g., pitch, tone, rhythm) on the millisecond level, and detects specific deviations in them that are indicative of a discrepancy between the executive’s understanding of the firm and the statements he/she is making about it. Our data predict some aspects of the immediate reaction of the market (e.g., immediate volatility r=0.10), the reaction of the analysts (in particular the earnings surprise, r=0.07), as well as several future performance indicators of the firm (e.g., ROE r=0.15, Tobin’s q r=0.18, debt/asset ratio r=0.10). API live data is available for clients on a 4-hour delay post earnings call. This research comes from a historical dataset covering 63,126 audio recordings of earnings calls (2005 – 2020) from 1,227 different equities that represent the Russell 1000, delivery of 104 pre-processed voice features. There is no survivorship bias in the data. Russell 3000 coverage also available upon request.