Frequent Subtree-Based Event Language Expression Pattern Discovery

Event language expression plays an important role in the event ontology. Event language expression patterns are the part of the event language expression, which indicate the customary habits when human describe the events in natural language. They are implicit in texts and can be represented by frequent subtrees. For the sentences that describe the same event class in CEC 2.0, firstly, we did semantic dependency parsing for each sentence and got the semantic dependency graphs by LTP, then transform them to semantic dependency trees, and mined frequent subtrees from these semantic dependency trees by the PETreeMiner algorithm. We verified the validity of the mined frequent subtrees through detecting event elements in texts with them. The value of F1 is 76.4%.