Joint Entity and Relation Extraction for Free Text

Both entity and relation extraction can benefit from being performed jointly, allowing each task to correct the errors of the other. The syntactic knowledge extraction is specific knowledge extraction, which automatically extracts the characteristic words and patterns based on hierarchy bootstrapping machine learning. A new method for joint entity and relation extraction using a graph called as a “card-pyramid”. This graph compactly encodes all possible entities and relations in a sentence, reducing the task of their joint extraction to jointly labeling its nodes. An efficient labeling algorithm that is analogous to parsing using dynamic programming. The experiment demonstrates the benefit of joint entity and relation extraction.