Ontology field guide
By NOSIBLE Research
Ekman-7 Emotion Ontology
Ekman-7 separates seven reported emotional frames that one sentiment score collapses and preserves them for direct comparison through time.
Positive–negative sentiment hides information. Fear signals threat. Anger signals blame. Sadness signals loss. Disgust signals rejection. Ekman-7 assigns each event one dominant frame: anger, contempt, disgust, fear, happiness, sadness or surprise. The categories come from Ekman’s basic-emotion research. World applies them to reporting, not individual psychology or private emotional states.[1][2]
World applies Ekman-7 to company, government, conflict, hazard, policy and science events. Researchers can test whether each emotion relates to volatility, capital flows, credit spreads or later events. This requires point-in-time data, a defined investment universe, event window and matched baseline. This guide provides definitions, base rates, representative events and downloadable data needed to evaluate the feature before modelling.
- Categories
- 7 categories
- Structure
- Flat, no hierarchy
- World events labelled
- 83.3%
- Stable codes
- Since v1
- Release data
- CC0
Foundations
Ekman-7 distinguishes reported emotional frames without inferring the private emotions people feel
Ekman treats emotions as distinct response systems, not points on a single positive–negative scale. Each category combines signals, appraisal, context and likely action. Anger indicates obstruction or offence. Fear indicates threat. Disgust indicates rejection. Sadness indicates loss. Happiness indicates desirable outcomes. Surprise indicates new information. Cross-cultural studies support shared category recognition, but expression and interpretation vary by culture. The ontology is a testable classification system, not a tool for inferring private psychological states.[1][2][3]
World classifies the dominant emotional frame expressed in event reporting. It does not estimate how an author, source or market participant feels. One event may contain several frames, and repeated coverage can increase counts without creating another economic shock. Researchers should test the seven categories against polarity and richer representations. Results should remain stable across sources, sectors, languages, regions and time. Any incremental performance should be linked to specific emotions and market regimes, not attributed to the ontology itself.[4]
Categories
Ekman-7 assigns seven independent emotion labels through a predictable, scalable event classification
Ekman-7 is flat: every event receives one label. Anger and contempt are peers, like fear and sadness. Each label describes the dominant frame in reporting, not anyone’s internal state. The explorer shows each category’s definition, code, World V2-preview count and share of labelled events and the full corpus. Use the signals as hypotheses: fear can isolate threat, anger blame, surprise new information and disgust rejection. Search supports ontologies with thousands of categories.
Ontology index
14,083,544 of 16,908,539 World events carry Ekman-7 Emotion Ontology labels.
anger
anger
Definition
Reporting framed by anger through blame, confrontation, demands for redress or responses to offence or blocked goals.
Research use
Blame, confrontation and redress
- Code
- anger
- Events with label
- 2,548,854
- Share of labelled
- 18.1%
- Share of World
- 15.1%
Events with label is the selected count. Label share divides it by 14,083,544 assigned events; World share divides it by all 16,908,539 events.
Representative World V2-preview events
One strong classified example per available year, with up to ten years shown.
Iceland President Defends Icesave Veto Amid UK and Dutch Anger
Coverage 5Egypt Prosecutor Appeals Mubarak Acquittal Amid Public Anger
Coverage 112Kanye West Sentenced to Probation and Anger Management for Paparazzi Assault
Coverage 59Obama Debuts Anger Translator Keegan-Michael Key at Correspondents Dinner
Coverage 46Key and Peele Return for Final Obama Anger Translator Sketch on Daily Show
Coverage 92Hong Kong Student Dies in Fall During Protests, Sparking Anger
Coverage 254Newsom Fights Partisan Recall Amid Pandemic and Unemployment Anger
Coverage 60Xi Jinping Maintains Strict Zero-Covid Policy Amid Public Anger
Coverage 117Sarfaraz Khan Debut Ends in Run Out as Rohit Sharma Throws Cap in Anger
Coverage 895Rand Paul Confronts Trump's DHS Pick Mullin Over Anger Issues
Coverage 6,756
Browse categories to compare definitions, World V2-preview statistics and representative classified events.
Case-study dataset: World V1.2. These historical studies retain their original event classifications and cohort statistics. The category explorer and downloads on this page use World V2-preview.
Example usage 1 of 3
Example Usage: FTX coverage shifts from fear during collapse to anger during prosecution
World V1.2 contains 690 Ekman-labelled events tagged with FTX or Sam Bankman-Fried. During the November–December 2022 collapse, fear accounts for 55.3% of 217 events and anger 31.3%. Across the trial window, anger rises to 71.2% while fear falls to 21.6%. Around sentencing, anger reaches 88.9% and fear 7.4%.[5][6][7]
Collapse reporting emphasizes threat, insolvency and contagion. Trial and sentencing emphasize blame, wrongdoing and punishment. A 14-day decay preserves the transition without treating each observation as a new regime. This exposes financial-risk mechanisms that a single negative sentiment score cannot distinguish across reported phases in World.[6][7]
- Collapse · 217 events
- Fear 55.3% · Anger 31.3%
- Trial · 111 events
- Anger 71.2% · Fear 21.6%
- Sentencing · 27 events
- Anger 88.9% · Fear 7.4%
The decayed series use a 14-day half-life and a seven-event prior estimated only from observations before October 2022. Later observations cannot change the baseline. The longer memory suppresses isolated daily reversals while preserving the collapse, trial and sentencing transitions. This describes media coverage. It does not test returns, causality, attribution or economic impact on exposed assets.
Example usage 2 of 3
Example Usage: Election day creates a brief happiness spike before anger quickly returns
The 2024 United States presidential election produces a sharp but temporary change in classified emotion. Happiness reaches 23.9% in the seven days beginning after election day, up from 6.3% during the prior week. Anger falls from 52.4% to 40.6%, while surprise rises as the outcome becomes known.[9]
The transition reverses within two weeks. Happiness falls below 11% and anger returns above 50%, so the improvement has a short half-life. Emotion categories separate outcome confirmation from renewed political conflict. Researchers can then test policy attention, sector exposure, volatility and event-signal decay against each phase.[10]
- Happiness
- +17.6pp
- Anger
- -11.8pp
- Surprise
- +5.3pp
The fixed English cohort covers 15 October through 25 November 2024 and requires election language plus a presidential candidate or voting term. Each line pools Ekman counts across the current and prior six days, then divides by all Ekman-assigned cohort events in that window.
Example usage 3 of 3
Example Usage: COVID emotion states reveal distinct crisis phases hidden inside negative sentiment
COVID coverage doubles after the World Health Organization declares a pandemic on 11 March 2020. Normalized Ekman-labelled attention rises from 50,374 to 102,045 events per million across matched fourteen-day windows. Disgust remains the largest state, but its share falls from 49.3% to 34.9% as the reporting agenda broadens beyond contamination.[8]
Anger rises from 6.4% to 16.6%, while sadness triples from 4.5% to 13.5%. Fear falls despite the wider crisis. Those states map to different mechanisms: contamination, policy conflict and realized loss. Researchers can condition volatility, sector exposure and policy-response tests on crisis phase instead of treating every negative COVID event as equivalent.[1]
- Normalized coverage
- +51,671 per million
- Disgust
- -14.4pp
- Anger
- +10.2pp
- Sadness
- +9.0pp
The fixed English COVID cohort compares 26 February to 10 March 2020 with 11 to 24 March. Each line pools emotion counts across the current and prior thirteen days, then divides by all Ekman-assigned cohort events in that window. The marker is the World Health Organization's pandemic declaration.
Data and sources
Download Ekman-7 and References
Downloads
Download categories and counts as CSV, or the complete machine-readable release as JSON.
Release details and citation files
ManifestRelease READMECC0 license and scopeCitation fileChangelog
Need the complete World event schema? Open the World data dictionary.
References
- [1]Ekman, P. (1992). An argument for basic emotions. Cognition & Emotion, 6(3-4), 169-200.
- [2]Ekman, P., & Friesen, W. V. (1971). Constants across cultures in the face and emotion. Journal of Personality and Social Psychology, 17(2), 124–129.
- [3]Ekman, P. (1992). Facial expressions of emotion: New findings, new questions. Psychological Science, 3(1), 34–38.
- [4]Hutto, D. D., Robertson, I., & Kirchhoff, M. D. (2018). A new, better BET: Rescuing and revising Basic Emotion Theory. Frontiers in Psychology, 9, 1217.
- [5]FTX Trading Ltd. (2022, November 11). Voluntary Chapter 11 petitions filed in the United States Bankruptcy Court for the District of Delaware.
- [6]U.S. Department of Justice. (2023, November 2). Statement on the guilty verdict in the trial of Sam Bankman-Fried.
- [7]U.S. Attorney's Office, Southern District of New York. (2024, March 28). Samuel Bankman-Fried sentenced to 25 years in prison.
- [9]Federal Election Commission. Official 2024 presidential general election results.
- [10]Funck, A. S. (2024). A meta-analytic assessment of the effects of emotions on political information search and decision-making. American Journal of Political Science.
- [8]World Health Organization. (2020). Director-General's opening remarks at the media briefing on COVID-19, 11 March 2020.
Continue exploring