Extraction of Topic and Event Keywords from News Story

Abstract

The topic/event related keywords, i.e. key-verbs and key-nouns, identification in news stories always dominates the performance of the news processing. However, little literature has been published on the key-verbs and key-nouns identification in news stories. This paper proposes a topic/event detection method that exploits the characteristics of news writing and the grammar properties of language to identify the topic and event keywords that adequately capture the topical information of the news stories for improving the performance of the automatic news processing, such as news classification, topic detection and tracking, retrieval and summarization, etc. We apply a news clustering system to evaluate the effectiveness of the topic/event related keywords exaction approach. Experimental results show that the proposed method can extract commendably accurate topic and event keywords to represent the news and can efficiently produce news clustering with higher quality compared with the news clustering without topic/event detection processing.