The Design And Implementation Of Domain-specific User-Driven Automatic Abstracting System

In order to benefit from the vast amounts of information quickly and accurately find the needed information, topics on automatic abstracting have become more and more popular.Compared with the traditional automatic abstracting methods, we adopted the statistic methods to develop user-driven automatic abstracting systems.Referring is a common phenomenon in the real life. In the process of papers'handling, referring often leads to inaccurate results. To solve the problem caused by the referring, combined with the current studies, this paper introduced resolution algorithm and the implementation of domain-specific user-driven automatic abstracting system based on resolution algorithm. By analyzing the referring phenomenon in the original text, term frequency and sentence importance are recalculated to obtain new abstracting results.This paper describes the implementation of the user-driven automatic abstracting system in specific domain, and the experiment results verified the improvement caused by the resolution algorithm.In order to verify the feasibility and effectiveness of the method in this paper, the internal evaluation method is adopted to evaluate the abstract system. 100 papers in political subjects were taken from the library to evaluate abstracts made by our system. According to the evaluation result, we can see that the system based on our algorithm is more effective than the standard system.