CCKB: Constructing and Evaluating a Large-Scale Causal Knowledge Base for Chinese

Causal knowledge outlines the logical progression of events in nature and society, which is crucial for AI applications like intelligent Q&A and text comprehension. Currently, the research on the construction of a causal knowledge base is relatively mature only in English, and there are few studies in the Chinese field, especially the lack of practical large-scale Chinese causal knowledge bases. This paper introduces a novel algorithm for Chinese causal pattern discovery, which leverages bilingual alignment and the identified English causal patterns to identify the causes, effects, and causality indicators in Chinese text. We then collects causal events from a Chinese corpus of about 500GB and perform operations such as deduplication, filtering, desensitization, and vectorization to form a Chinese causal knowledge base (CCKB) of 10 million pieces. Evaluations indicate the constructed CCKB reaches an accuracy of 88.6%, which is available for public access via https://github.com/Ewillingfly/ChineseCausalityKB.