PatentMiner: Patent Vacancy Mining via Context-Enhanced and Knowledge-Guided Graph Attention

Although there are a small number of work to conduct patent research by building knowledge graph, but without constructing graph using documents and combining latest natural language processing methods mine hidden rich semantic relationships in existing patents predict new possible patents. In this paper, we propose vacancy prediction approach named PatentMiner potential based on (KG) attention mechanism. Firstly, over time (e.g. year) is constructed carrying out entity recognition relation extrac-tion from documents. Secondly, Common Neighbor Method (CNM), Graph Attention Networks (GAT) Context-enhanced (CGAT) proposed perform link the dig triples. Finally, defined means co-occurrence relationship, that is, each represented as fully connected subgraph containing all its entities graph; Furthermore, task which predicts with newly added links pa-tent. The experimental results demonstrate our predic-tion can correctly much better than baseline. Meanwhile, still has significant room im-prove.