A domain-based automatic text summarization system

A new method for automatic text summarization which combines sentence extraction and domain knowledge is proposed, and an automatic text summarization system is developed based on the above method. First, construct a corpus and knowledge base based on domain knowledge. And every sentence in the knowledge base is expressed by a vector of eight elements including domain word features, sentence location features, sentence length feature, associated features in knowledge base and etc. Next, calculate the weight of each sentence to compose the coarse text summary by selecting the sentences with bigger weights. Finally, post-process the coarse text summarization to create a smooth and readable summary infilling the text summarization frame based on grammar and domain knowledge base. Domain experts' evaluation on our system shows that the text summarization method presented in this paper is effective and feasible.