Inflation Narratives from a Machine Learning Perspective

Inflation narratives explain inflation changes and affect expectations. Manu- ally identifying them is cumbersome, prompting the need for scalable algo- rithms. Narratives comprise events, causal relations, and arguments, repre- sented as graphs with event and argument nodes. Causal relations indicate cause-and-effect relationships between events using directed edges. Our main objective is to extract narratives from text to enhance a knowledge graph for analysis like social network analysis or edge prediction. We address two sub- problems: event extraction, involving event type and argument identification, and event deduplication. Second, we extract causal relations as expressed by authors, not necessarily true causal links between events in the text.