Rules Used in Neural Rule Engine

We release the rules served for Relation Classification (SemEval-2010 Task 8) in the paper:

Shen Li, Hengru Xu, Zhengdong Lu, Generalize Symbolic Knowledge With Neural Rule Engine

Each rule is defined as: [positive]@@[negative]

where [positive] means something that a case must have if the case is matched by the rule, and [negative] means the case mustn't have.

For example, given a case and a rule:

Case: The <e1>island</e1> was farmed for <e2>oats</e2> and barley until 1796 when the owner mandated that oak and alder trees be planted.

Rule: island.*</e1>.*farmed for.*<e2>@@<e2>(year|profit).*</e2>

since the case has "island.*</e1>.*farmed for.*<e2>" and hasn't "<e2>(year|profit).*</e2>", the case is matched by the rule.