Study On The Technology And Application Of Biased Summarization

Corresponding with the rapid development of the Internet, we are surrounded by an immensesea of information. How to get accurate and valid information from this vast information sea is animportant work in information extraction. A specific search engine can not satisfy the actualapplications so far.Automatic summarization was first studied in Luhn's paper "The Automatic Creation ofLiterature Abstracts" in 1958 and has been a hot research issue in the field. Automatic textsummarization should not only be simple, concise, accurate, and efficient as text summarizations,but also can deal with mass data quickly. It has become an important tool for people to obtaininformation quickly. During the past fifty years, many progresses have been made in genericsummarization. And the biased summarization has attracts many researchers in recent years,because it makes the re-treat of information more easy and plays an important role in InformationResearch (IR), Question and Answer (Q & A) etc. The biased summarization is becoming a hot spotin automatic text summarizations.Automatic summary, common summary, and biased summary are discussed deeply in thepaper. After reviewing the history of biased summary, a practice biased summary system isdesigned and implemented based on the algorithm of the density distribution of keywords. Thissystem is developed in Visual C++6.0 and SQL Server 2000, and provides three automaticsummaries in different compress rate according to users' demand.On the study of the application of biased summary, two contrast experiments are made to testthe practicability and the effectiveness of the system on IR and Q & A tasks. The experiment resultsshow, the system satisfies the demand of searching some special information. And the averagecorrect rates of the test are 72.5% and 86.5% respectively.Finally, an evaluation method is proposed based on hybrid strategy. And the correctness andvalidity of this method is tested by two different experiments. The test result comes up to what weexpect. It is also demonstrated that the quality of summaries created by the system based on thedensity distribution of keywords is better than those created on the similarity of sentence.