Keyword Extraction Method for Complex Nodes Based on TextRank Algorithm
Qingyun Zhou, Yuansheng Fang, Shang Zhenlei, Zhong Wanli · 2020 International Conference on Computer Engineering and Application (ICCEA) · 2020
Keywords are important way for people to quickly understand content of document and grasp subject, and keyword extraction technology is significant way to quickly obtain core meaning of text information, which has wide range of applications in fields such as intelligence, journalism, information retrieval and natural language understanding. However, traditional TextRank algorithm refers to local co-occurrence relationship among text words, which does not pay much attention to complex network structure characteristics of word graphs. Therefore, structure of network is adjusted by removing nodes to separate sub-networks with layers. Moreover, taking into account complex network structure characteristics of word nodes, method of word node removal is introduced as well. Meanwhile, value of sliding window is increased so that ranking can be obtained through multiple iterations, and then one of the highest ranking keyword nodes will be removed in turn. Besides it, Keyword extraction is then performed on each subtopic where sub-keywords are determined based on ranking of candidate keywords, and key nodes of text network are added to sequence so that keyword extraction can be achieved, which achieves improvement of traditional TextRank algorithm, and accuracy, recall, and F value are all improved as well.