NER‐RAKE: An improved rapid automatic keyword extraction method for scientific literatures based on named entity recognition

Abstract

AbstractThe proposing of an accurate and efficient model for automatic keyword extraction of scientific literature is conducive to promoting the development of academic text mining research, such as literature retrieval optimization, research topic discovery, topic evolution analysis, emerging trend detection and so on, with the continuous expansion of digital academic resources. A keyword extraction method is proposed called NER‐RAKE which combines Named Entity Recognition (NER) process with Rapid automatic keyword extraction (RAKE), Bidirectional Long Short‐Term Memory Network Conditional Random Field (BiLSTM‐CRF) is used to recognize the domain entities in the scientific literature so as to enrich the list of candidate keywords divided by RAKE, and the frequency threshold is set to ensure the validity of candidate keywords. NER‐RAKE is better than RAKE in accuracy and recall rate through the keyword extraction experiment in the computer science field. It is effective to combine the process of NER with RAKE, which can improve the efficiency of keyword extraction.