// the find
liuhuanyong/ComplexEventExtraction
A concept and obvious expression pattern collection of Chinese compound event extraction which then be evolved into ComplexEventGraph，本项目提出了中文复合事件的概念与显式模式，包括条件事件、因果事件、顺承事件、反转事件等事件抽取，并形成事理图谱。
A Python project from a Chinese NLP researcher that extracts compound event relationships from Chinese text using explicit connective patterns — causal, conditional, sequential, and contrastive. It runs pattern matching over corpora to build what the author calls an 'event graph' (事理图谱). Aimed at Chinese NLP researchers doing knowledge graph or event reasoning work.
The taxonomy of compound event types (causal, conditional, contrastive, sequential) is clearly defined and practically motivated. The pattern list in complex_sentence.py is extensive and grounded in real Chinese connective usage, not hand-waved. Running this over 10M news articles and surfacing frequency stats for 237 patterns is useful empirical grounding. The README honestly shows the output rather than just promising it.
Abandoned in December 2018 — seven years ago — with no updates since. There is essentially no code in this repo beyond a single pattern file; the actual extraction pipeline, graph construction, and data are not here. Pure rule-based pattern matching over connectives will break badly on any text that uses implicit causality or non-standard phrasing, which is most of real Chinese writing. No tests, no evaluation metrics, no benchmark — just cherry-picked example outputs.