基于MO改进算法的多文档文摘句子排序

针对CO和MO两种文摘句排序方法的缺陷，提出了一种将局部主题间的内聚度和MO方法相结合进行文摘句排序的新方法。在对局部主题间的相对位置统计的基础上，建立它们间的关系有向图和计算彼此间的内聚度；排序过程中每从有向图中输出一个顶点时，从剩余顶点中查找与其具有最大内聚度的顶点，若该内聚度大于阈值，则将这两个顶点所代表的局部主题的文摘句置于摘要中相邻的位置。实验结果表明，该算法排序生成的文摘更具有连贯性和可读性。

Improved Sentence Ordering algorithm for Multi-Document Summarization based on Majority Ordering

The limitations of Chronological Ordering and Majority Ordering are introduced and a new sentence ordering algorithm is proposed in which the mutual cohesion between themes and the Majority Ordering are associated into the sentence ordering. Based on the statistical data about the relative position in each pair of themes, a directed graph about themes is built and the mutual cohesion is computed. When a vertex is outputted from directed graph, search a vertex from the remaining vertexes which has the greatest cohesion with the vertex. If the cohesion is bigger than the threshold value, the sentences from these two themes should be placed to the adjacent location in the summary. Experimental results show that summary generated by the algorithm is more coherent and more readable.