A novel keyphrase extraction method by combining FP-growth and LDA
Hao Sun, Bing Li, Bo Han · 2017
Fast-growing technologies like cloud-computing, big data, mobile Internet, artificial intelligence, etc. have driven the emergences of a lot of new phrases. In this paper, we propose a novel keyphrases extraction method with two steps by combining FP-growth algorithm and Latent Dirichlet Allocation (LDA) topic modeling. In the first step, we apply FP-growth algorithm to obtain frequent neighborhood words co-occurring frequently as candidate phrases. In the second step, we extract significant keyphrases by LDA models. Our experiments on two datasets CVE-2015 and 20-newsgroups have shown that the proposed approach can extract significant keyphrases and these phrases can help improve the text classification accuracy.