TEA Nets: AI framework models targets, events and actors in text

A new computational framework, known as Target-Event-Agent Networks (TEA Nets), has been developed by researchers to identify subjects (Agents), verbs (Events), and objects (Targets) within texts. This framework, grounded in cognitive network science and artificial intelligence, is available as an open-source Python library. Testing on the LOCO conspiracy corpus, which includes 4,227 texts, indicated that highly conspiratorial narratives linked personal pronouns ("I", "you", "we") to the same actions at twice the rate of low-similarity conspiracy narratives. Additionally, high-conspiracy narratives associated person-focused elements ("you", "people") with anger-inducing actions, surpassing the random baseline (z = 2.63, p < .05), a pattern not found in low-similarity texts. The framework also excelled in emotion detection, semantic frame analyses, and linguistic studies across conspiracy texts and responses generated by LLMs.