Combination of Unsupervised Keyphrase Extraction Algorithms

Abstract

Key phrase extraction plays a significant role in many language processing tasks such as text summarization, text categorization and information retrieval. However, none study combines several approaches to improve the performance of key phrase extraction. This paper first implements three representative unsupervised algorithms TfIdf, Text Rank and Expand Rank, and then proposes a generalized framework using serial, parallel and voting methods on combining these algorithms for comprehensive analysis of key phrase extraction. Experimental results, carried out on an evaluation dataset including 1040 abstracts from Chinese thesis, demonstrate the remarkable performance of some combination approaches.