Keyword Extraction Based on tf/idf for Chinese News Document

摘要

Keyword extraction is an important research topic ofinformation retrieval. This paper gave the specification of keywords in Chinese news documents based on analyzing linguisticcharacteristics of news documents and then proposed a new keyword extraction method based on tf/idf with multi-strategies. Theapproach selected candidate keywords of uni-, bi- and tri- grams,and then defines the features according to their morphologicalcharacters and context information. Moreover, the paper proposedseveral strategies to amend the incomplete words gotten from theword segmentation and found unknown potential keywords innews documents. Experimental results show that our proposedmethod can significantly outperform the baseline method. We alsoapplied it to retrospective event detection. Experimental resultsshow that the accuracy and efficiency of news retrospective eventdetection can be significantly improved.