Abstract

Event-based summarization and biography are involved with processing people information, but they still have some deficiencies in identifying key people. We present a method of identifying key people from a single document in this paper. After analyzing and quantifying three kinds of associative strengths between people and events, between people and between events, a people event map is constructed to represent a document, and the people's importance can be computed using the PageRank algorithm based on people event map. We apply two methods to identify key people from 834 documents. The average accuracy achieves 92.7% using the method based on people event map to identify key people, and the average accuracy is 80.2% by the method of counting the people's frequencies.

Paper Details