Single-Document Automatic Abstracting System Based on Topic Partition

The traditional single-document automatic abstracting based on statistical extracts a number of sentences sorted by the importance of the sentences to form summarization, which often neglects semi important topics of the text, and makes the summarization not completely. To overcome this shortcoming, the paper presents an improved k-means algorithm to divide topic in the analysis of text structure. Then proper sentences are extracted from every topic to form summary, which can enhance the summary´s information coverage and let the summary comprehensively. Compared with the abstracting based on statistic the result shows that the method based on topic partition obtains better effect.