COMPARISON OF THREE TEXT SUMMARIZATION METHODS

摘要

In this paper, we present three generic text summarization methods. Each method creates a text summary by ranking and extracting sentences from an original document. The first method, SUMMARIZER 1, uses standard information retrieval (IR) methods to rank sentences. The second method, SUMMARIZER 2, uses the Latent Semantic Analysis (LSA) technique to identify semantically important sentences, for summary creations. The third method, SUMMARIZER 3, simply uses the TF*IDF weighting scheme. Evaluations of the three methods are conducted using Document Understanding Conferences (DUC) datasets from NIST. We have compared the summary (extracts) of each method with the manual summaries (abstracts). Summarizer 3, with its lowest overhead, has comparable performance to summarizer 1. All analysis show that LSA technique, as used in SUMMARIZER 2, does improve text summarization. Summarizer 3 has the best performance on an average.