Collecting All Causal Knowledge

Summary

Causal knowledge is seen as one of the key ingredients to advance artificial intelligence. Yet, few knowledge bases comprise causal knowledge to date, possibly due to significant efforts required for validation. Notwithstanding this challenge, we compile CauseNet, a large-scale knowledge base of claimed causal relations between causal concepts. By extraction from different semi- and unstructured web sources, we collect more than 11 million causal relations with an estimated extraction precision of 83\% and construct the first large-scale and open-domain causality graph. We analyze the graph to gain insights about causal beliefs expressed on the web and we demonstrate its benefits in basic causal question answering. Future work may use the graph for causal reasoning, computational argumentation, multi-hop question answering, and more. Download We provide three versions of our causality graph CauseNet: Statistics Relations Concepts File Size CauseNet-Full 11,609,890 12,186,195 1.8GB CauseNet-Precision 199,806 80,223 135MB CauseNet-Sample 264 524 54KB Data Model The core of CauseNet consists of causal concepts which are connected by causal relations. Each causal relation has comprehensive provenance data on where and how it was extracted. Examples of Causal Relations Causal relations are represented as shown in the following example. Provenance data is omitted. { "causal_relation": { "cause": { "concept": "disease" }, "effect": { "concept": "death" } } } For CauseNet-Full and CauseNet-Precision, we include comprehensive provenance data. In the following, we give one example per source. For relations extracted from natural language sentences we provide: surface: the surface form of the sentence, i.e., the original string path_pattern: the linguistic path pattern used for extraction ClueWeb12 Sentences clueweb12_page_id: page id as provided in the ClueWeb12 corpus clueweb12_page_reference: page reference as provided in the ClueWeb12 corpus clueweb12_page_timestamp: page ...