Ever since Culturomics 2.0 demonstrated the predictive power of global-scale sentiment analysis in understanding global risk and launched the modern renaissance in latent emotion-based risk assessment, there has been immense interest in the ways in which emotion, especially dimensions beyond simple positive/negative scoring, can offer unprecedented hidden insights into the future courses of human society.
GDELT offers myriad options for assessing global emotion. Each day it scores every news article it monitors in its 65 live-translated languages across thousands of emotions through both its built-in GDELT Core Emotion score, designed to offer a robust generalized "tone" score and the GKG's GCAM suite, which brings together dozens of existing sentiment tools that catalog an incredible wealth of dimensions, from traditional measures like anxiety and fear of the future to new research like eMFD's moral foundations.
These can be used to trivially sort news coverage by emotion or aggregated at scale to literally map the emotion of the world's news media in realtime.
We also score a small random selection of English language coverage each day through a state of the art neural sentiment algorithm for comparison.
Using the new Web NGrams 3.0 dataset it is now possible to apply your own BOW or narrow-window contextualized statistical and neural sentiment algorithms directly across the more than 150 languages GDELT monitors each day. Applications range from automated training construction to model edge case identification to applying production models to assess risk and narrative emotional undercurrents in realtime.
We are tremendously excited to see the kinds of new sentiment research this powerful dataset makes possible and would love to hear from you with any interesting applications you build.