Generate Keyphrases by Combining Hidden Context and Word Context

Abstract

As the network develops, highly generalized indexing information becomes more and more important. Keyphrases are multiple words that can simply summarize the semantics of an article. Keyphrase generation is the basis for many natural language processing tasks. Although there are many tasks for keyword generation, they consider hidden layer information and almost ignore each independent word information in the source text. So we attach word-level information to the model and combine this information with hidden layer information. Experimental results show that our model can effectively improve the accuracy of the generated keyphrases.