A New Two-Stage Keyword Extraction Method for T.V. News Contents

Abstract

In this paper, we address a two-stage keyword extraction method for Internet Chinese T.V. news document. The first stage, a lexicon matching process, extracts most meaningful words from a document. The second stage performs a statistical analysis of the term frequency of each word in the document. We also propose an integrated algorithm to combine these two results to conclude a keywords list that can best outline or represent the original documents. We have implemented a prototype system to evaluate the functionality of the proposed method. Preliminary experimental results show that the proposed method seems very promising. We also implement this keyword extraction and indexing methods in a Internet TV news browsing system, http://nn.csie.nctu.edu.tw/Project1-1/introduction.htm. A user can give set of query keywords to retrieve preferred multimedia TV news stories.