A Spreading Activation Based Sentence Ordering Model Collaborated with Cognitive Logical Structure

Abstract

As novel web social Media emerges on the web, large-scale unordered sentences are springing up. Although these massive unordered sentences have rich information, loose semantic association and unordered distribution make users hard to order these sentences toward well semantic coherence manually. Automatic sentence ordering to maximize semantic coherence is a significantly practical problem. Some existing approaches have proposed some coherence measure methods of sentences, however some challenging issues have not been solved, which includes: 1) how to collaborate some basic cognitive logical structures into sentence ordering, 2) what is sound cognitive mechanism to guide sentence ordering. To solve these issues, we propose a spreading activation based sentences ordering model collaborated with cognitive logical structure. Our sentence ordering model collaborates with three basic cognitive logical structures, which includes summary is prior, illustrate follows, illustration is prior, summary follows and illustration is one by one. Besides, our sentence ordering model pursues coherence sentence order under guide by spreading activation, which actives knowledge from most associated keywords to most associated sentences. To validate correctness of our model, some experiments are conducted to measure the accuracy of the automatically generated sentence order. The results show our model can give sentences order with higher accuracy in less iteration. The sentence ordering model can be applied in automatic text organization and summarization.