The early works attempted to extract causal relations using knowledge-based inference technologies (Kaplan, 1991) These studies were based on hand-coded, domain-specific knowledge bases which are difficult to scale up for realistic applications. Recently there has been increasing interest in automatic causality extraction from texts, which can be classified into two approaches: the pattern-based approach and statistic-based approach.
Existing statistical methods for causality acquisition used one or more distribution characteristics of two events in the text. These major features are: (1) Co-Occurrence feature: the cause event and effect one may co-occur frequently; (2) Object-Sharing feature: the related two events may share a common participant; (3) Temporal feature: the cause event occur before (or simultaneously with) the effect event; (4) Distance feature: the two events may appear inside locally coherent text (in the same sentence particularly).